What changed in AI model pricing

Providers change rates, add and drop models, and quietly switch to cheaper precision without announcing any of it. We snapshot every host daily and publish the differences.

Last checked 12 October 2026RSS feed

2026-10-12

107 changes since 2026-10-09

withdrawnQwen3 VL 32B Instruct is no longer offered by any host we track.
withdrawnQwen3 VL 8B Thinking is no longer offered by any host we track.
withdrawnQwen3 8B is no longer offered by any host we track.
withdrawnQwen3.6 Max Preview is no longer offered by any host we track.
withdrawnQwen3 235B A22B is no longer offered by any host we track.
withdrawnQwen3 Coder Plus is no longer offered by any host we track.
withdrawnQwen3 Max is no longer offered by any host we track.
withdrawnQwen3 VL 235B A22B Thinking is no longer offered by any host we track.
withdrawnQwen Plus 0728 is no longer offered by any host we track.
withdrawnQwen3 30B A3B Thinking 2507 is no longer offered by any host we track.
withdrawnQwen3 Max Thinking is no longer offered by any host we track.
price cutGLM 5.3 Flash input price fell 85%, $0.150 to $0.023 per million tokens.
price riseGLM 5.3 Flash output price rose 99%, $0.500 to $1.20 per million tokens.
hostsGLM 5.3 Flash is now served by 32 hosts, down from 33.
price riseQwen3 VL 8B Instruct input price rose 99%, $0.117 to $0.250 per million tokens.
price riseQwen3 VL 8B Instruct output price rose 65%, $0.455 to $0.750 per million tokens.
hostsQwen3 VL 8B Instruct is now served by 1 hosts, down from 2.
contextQwen3 VL 8B Instruct context most hosts serve moved from 131K to 262K.
hostsQwen3 Coder 30B A3B Instruct is now served by 2 hosts, down from 3.
hostsGLM 5.3 Flash (batch) is now served by 32 hosts, down from 33.
hostsQwen3.7 Flash is now served by 2 hosts, up from 1.
hostsQwen3 VL 30B A3B Instruct is now served by 3 hosts, down from 4.
contextQwen3 VL 30B A3B Instruct context most hosts serve moved from 131K to 262K.
hostsgpt-oss-120b (batch) is now served by 24 hosts, up from 23.
best valueGPT-6.1 Sol (batch) became best value — nothing cheaper now matches it.
hostsQwen3 235B A22B Instruct 2507 is now served by 7 hosts, down from 9.
price riseKimi K2.5 input price rose 11%, $0.450 to $0.500 per million tokens.
price riseKimi K2.5 output price rose 11%, $2.25 to $2.50 per million tokens.
hostsQwen3 14B is now served by 2 hosts, down from 3.
hostsQwen3.8 Max (0902) is now served by 2 hosts, up from 1.
price riseKimi K2.6 input price rose 40%, $0.465 to $0.650 per million tokens.
price riseKimi K2.6 output price rose 39%, $2.45 to $3.41 per million tokens.
hostsQwen3 30B A3B is now served by 1 hosts, down from 2.
price riseDeepSeek V3.2 output price rose 90%, $0.420 to $0.800 per million tokens.
hostsDeepSeek V3.2 is now served by 12 hosts, down from 13.
hostsQwen3 Next 80B A3B Thinking is now served by 1 hosts, down from 2.
price riseDeepSeek Flash Latest input price rose 99%, $0.016 to $0.034 per million tokens.
price cutDeepSeek V4 Flash Latest input price fell 11%, $0.014 to $0.013 per million tokens.
price riseDeepSeek V4 Flash Latest output price rose 99%, $0.121 to $0.768 per million tokens.
price cutDeepSeek V4 Flash 0731 input price fell 11%, $0.018 to $0.016 per million tokens.
price cutDeepSeek V4 Flash 0731 output price fell 50%, $1.28 to $0.640 per million tokens.
hostsDeepSeek V4 Flash 0731 is now served by 27 hosts, up from 26.
price cutQwen3.8 27B input price fell 91%, $0.425 to $0.040 per million tokens.
price cutQwen3.8 27B output price fell 10%, $2.55 to $2.30 per million tokens.
hostsQwen3.8 27B is now served by 20 hosts, up from 18.
price riseNemotron 3.5 Lightning input price rose 49%, $0.047 to $0.070 per million tokens.
price riseNemotron 3.5 Lightning output price rose 49%, $0.134 to $0.200 per million tokens.
hostsNemotron 3.5 Lightning is now served by 5 hosts, down from 6.
hostsGLM 5.3 (batch) is now served by 41 hosts, down from 42.
price cutGLM Flash Latest input price fell 28%, $0.032 to $0.023 per million tokens.
price cutGLM Flash Latest output price fell 7%, $1.29 to $1.20 per million tokens.
hostsQwen3 VL 235B A22B Instruct is now served by 3 hosts, down from 4.
contextQwen3 VL 235B A22B Instruct context most hosts serve moved from 131K to 262K.
price cutQwen3.6 35B A3B input price fell 33%, $0.150 to $0.100 per million tokens.
price cutQwen3.6 35B A3B output price fell 5%, $1.00 to $0.950 per million tokens.
price riseNemotron 3 Ultra input price rose 20%, $0.500 to $0.600 per million tokens.
price riseNemotron 3 Ultra output price rose 9%, $2.20 to $2.40 per million tokens.
hostsgpt-oss-120b is now served by 24 hosts, up from 23.
hostsQwen3.8 Flash is now served by 2 hosts, up from 1.
hostsQwen3 Coder 480B A35B is now served by 3 hosts, down from 4.
hostsQwen3.7 Plus is now served by 2 hosts, up from 1.
price riseKimi K3 input price rose 46%, $0.500 to $0.730 per million tokens.
price cutKimi K3 output price fell 17%, $12.00 to $10.00 per million tokens.
hostsMuse Glimmer 30B is now served by 2 hosts, down from 3.
hostsQwen3 Coder Next is now served by 1 hosts, down from 3.
price riseDeepSeek Pro Latest input price rose 26%, $0.198 to $0.250 per million tokens.
price cutDeepSeek Pro Latest output price fell 38%, $8.00 to $5.00 per million tokens.
hostsLing 3.0 Flash Fin is now served by 1 hosts, down from 2.
price riseQwen3 235B A22B Thinking 2507 input price rose 96%, $0.230 to $0.450 per million tokens.
price riseQwen3 235B A22B Thinking 2507 output price rose 52%, $2.30 to $3.50 per million tokens.
hostsQwen3 235B A22B Thinking 2507 is now served by 1 hosts, down from 2.
contextQwen3 235B A22B Thinking 2507 context most hosts serve moved from 131K to 128K.
price cutQwen3.6 27B input price fell 29%, $0.450 to $0.320 per million tokens.
price riseQwen3.6 27B output price rose 19%, $2.70 to $3.20 per million tokens.
price riseGLM 4.6 input price rose 16%, $0.430 to $0.500 per million tokens.
price riseGLM 4.6 output price rose 14%, $1.75 to $2.00 per million tokens.
hostsQwen3 30B A3B Instruct 2507 is now served by 3 hosts, down from 5.
hostsQwen3.7 Max is now served by 2 hosts, up from 1.
price cutGLM 5.2 input price fell 29%, $0.084 to $0.060 per million tokens.
price cutGLM 5.2 output price fell 12%, $8.00 to $7.00 per million tokens.
price riseGLM 5.3 input price rose 99%, $0.039 to $0.079 per million tokens.
price cutGLM 5.3 output price fell 31%, $7.00 to $4.80 per million tokens.
hostsGLM 5.3 is now served by 41 hosts, down from 42.
hostsQwen3 Next 80B A3B Instruct is now served by 3 hosts, down from 4.
price riseQwen3 VL 30B A3B Thinking input price rose 45%, $0.200 to $0.290 per million tokens.
price cutQwen3 VL 30B A3B Thinking output price fell 58%, $2.40 to $1.00 per million tokens.
hostsQwen3 VL 30B A3B Thinking is now served by 1 hosts, down from 2.
contextQwen3 VL 30B A3B Thinking context most hosts serve moved from 131K to 262K.
price riseDeepSeek V4 Flash 0423 input price rose 99%, $0.007 to $0.030 per million tokens.
hostsDeepSeek V4 Flash 0423 is now served by 16 hosts, down from 17.
price riseKimi Latest input price rose 47%, $0.490 to $0.720 per million tokens.
price riseKimi Latest output price rose 85%, $13.00 to $24.00 per million tokens.
price riseGLM Latest input price rose 76%, $0.031 to $0.055 per million tokens.
price riseGLM Latest output price rose 14%, $4.40 to $5.00 per million tokens.
new hostTenstorrent started serving GLM 5.3 at $0.600 per million.
host leftNebius stopped serving GLM 5.3.
host leftRelace stopped serving GLM 5.3.
host leftRelace stopped serving GLM 5.2.
host leftDekaLLM stopped serving GLM 5.3 Flash.
precisionBaseTen switched DeepSeek V4.1 Flash from fp8 to fp32.
host leftDekaLLM stopped serving DeepSeek V4.1 Flash.
new hostVenice started serving gpt-oss-120b at $0.030 per million.
new hostNear AI started serving Qwen3.8 27B at $0.040 per million.
new hostTenstorrent started serving Qwen3.8 27B at $0.040 per million.
host leftRelace stopped serving DeepSeek V4 Pro 0813.
host leftNovita stopped serving DeepSeek V4 Flash 0423.
host leftRelace stopped serving DeepSeek V4 Pro 0423.

2026-10-09

131 changes since 2026-10-05

newClaude Haiku 5.5 (batch) appeared, served by 10 host(s) at $0.050 per million input tokens.
newNano Banana 2.1 appeared, served by 1 host(s) at $1.50 per million input tokens.
newGPT-6.1 Sol Pro (batch) appeared, served by 7 host(s) at $1.00 per million input tokens.
newStep 5 Preview appeared, served by 1 host(s) at $1.00 per million input tokens.
newClaude Haiku 5.5 appeared, served by 10 host(s) at $0.105 per million input tokens.
newGPT-6.1 Sol (batch) appeared, served by 8 host(s) at $1.00 per million input tokens.
newLing 3.0 Flash Sante appeared, served by 1 host(s) at $0.042 per million input tokens.
newMistral Large 4 appeared, served by 3 host(s) at $0.680 per million input tokens.
withdrawnKAT-Coder-Pro V2.5 is no longer offered by any host we track.
withdrawnERNIE 4.5 VL 424B A47B is no longer offered by any host we track.
hostsMiniMax M2 is now served by 2 hosts, down from 3.
hostsGLM 4.7 is now served by 6 hosts, down from 7.
hostsQwen3 Coder 30B A3B Instruct is now served by 3 hosts, down from 4.
hostsLlama 3.1 Euryale 70B v2.2 is now served by 1 hosts, down from 2.
price riseQwen3 235B A22B Instruct 2507 input price rose 3%, $0.087 to $0.090 per million tokens.
price riseQwen3 235B A22B Instruct 2507 output price rose 57%, $0.350 to $0.550 per million tokens.
hostsQwen3 235B A22B Instruct 2507 is now served by 9 hosts, down from 10.
hostsMiniMax M2.7 is now served by 6 hosts, down from 7.
price riseDeepSeek V3 input price rose 24%, $0.257 to $0.320 per million tokens.
price cutDeepSeek V3 output price fell 13%, $1.03 to $0.890 per million tokens.
hostsDeepSeek V3.1 Terminus is now served by 2 hosts, down from 3.
hostsDeepSeek V4 Flash Vision Exp is now served by 3 hosts, down from 4.
price cutQwen3.5-35B-A3B input price fell 47%, $0.150 to $0.080 per million tokens.
price cutQwen3.5-35B-A3B output price fell 25%, $1.00 to $0.750 per million tokens.
hostsMiniMax M2.1 is now served by 2 hosts, down from 3.
price cutKimi K2.6 input price fell 51%, $0.950 to $0.465 per million tokens.
price cutKimi K2.6 output price fell 39%, $4.00 to $2.45 per million tokens.
hostsKimi K2.6 is now served by 17 hosts, down from 18.
hostsKimi K3 (batch) is now served by 25 hosts, up from 24.
price riseDeepSeek V4 Pro 0813 input price rose 99%, $0.450 to $1.32 per million tokens.
price cutDeepSeek V4 Pro 0813 output price fell 21%, $5.00 to $3.96 per million tokens.
hostsDeepSeek V4 Pro 0813 is now served by 21 hosts, up from 20.
hostsHy3 is now served by 5 hosts, down from 6.
price cutDeepSeek V3.2 input price fell 8%, $0.280 to $0.259 per million tokens.
price riseDeepSeek Flash Latest input price rose 99%, $0.003 to $0.016 per million tokens.
price cutDeepSeek Flash Latest output price fell 50%, $2.40 to $1.20 per million tokens.
price riseLlama 3.3 70B Instruct input price rose 99%, $0.100 to $0.220 per million tokens.
price riseLlama 3.3 70B Instruct output price rose 56%, $0.320 to $0.500 per million tokens.
price cutDeepSeek V4 Flash Latest input price fell 5%, $0.015 to $0.014 per million tokens.
price cutDeepSeek V4 Flash Latest output price fell 91%, $1.28 to $0.121 per million tokens.
price riseDeepSeek V4 Flash 0731 input price rose 18%, $0.015 to $0.018 per million tokens.
hostsDeepSeek V4 Flash 0731 is now served by 26 hosts, down from 28.
hostsDeepSeek V4.1 Flash is now served by 31 hosts, up from 30.
hostsQwen3.8 27B is now served by 18 hosts, up from 17.
price cutNemotron 3.5 Lightning input price fell 22%, $0.060 to $0.047 per million tokens.
price cutNemotron 3.5 Lightning output price fell 16%, $0.160 to $0.134 per million tokens.
hostsNemotron 3.5 Lightning is now served by 6 hosts, up from 5.
hostsgpt-oss-20b (batch) is now served by 11 hosts, down from 12.
hostsGPT-6.1 Sol Pro is now served by 7 hosts, up from 6.
hostsGLM 5.3 (batch) is now served by 42 hosts, up from 41.
price cutGLM Flash Latest input price fell 9%, $0.035 to $0.032 per million tokens.
price riseGLM Flash Latest output price rose 99%, $0.500 to $1.29 per million tokens.
hostsQwen2.5 72B Instruct is now served by 1 hosts, down from 2.
hostsQwen3 VL 235B A22B Instruct is now served by 4 hosts, down from 5.
lost best valueMuse Spark 1.3 lost best value; something cheaper now matches its score.
hostsQwen3.6 35B A3B is now served by 10 hosts, up from 9.
best valueClaude Opus 5.5 (batch) became best value — nothing cheaper now matches it.
hostsDeepSeek V4.1 Flash (batch) is now served by 31 hosts, up from 30.
hostsMiMo-V2.6-Flash is now served by 8 hosts, up from 7.
hostsGPT-6.1 Sol is now served by 8 hosts, up from 7.
price riseInkling input price rose 5%, $0.950 to $1.00 per million tokens.
hostsgpt-oss-20b is now served by 11 hosts, down from 12.
hostsQwen3 Coder 480B A35B is now served by 4 hosts, down from 5.
price cutKimi K3 input price fell 25%, $0.670 to $0.500 per million tokens.
price cutKimi K3 output price fell 14%, $14.00 to $12.00 per million tokens.
hostsKimi K3 is now served by 25 hosts, up from 24.
price cutMuse Glimmer 30B input price fell 14%, $0.350 to $0.300 per million tokens.
price cutMuse Glimmer 30B output price fell 20%, $1.50 to $1.20 per million tokens.
price riseQwen3.5-27B input price rose 33%, $0.195 to $0.260 per million tokens.
price riseQwen3.5-27B output price rose 67%, $1.56 to $2.60 per million tokens.
hostsGemma 3 27B is now served by 5 hosts, up from 4.
hostsQwen3 Coder Next is now served by 3 hosts, down from 4.
price riseDeepSeek V4 Pro 0423 input price rose 99%, $0.209 to $0.955 per million tokens.
price riseDeepSeek V4 Pro 0423 output price rose 99%, $0.418 to $1.91 per million tokens.
hostsDeepSeek V4 Pro 0423 is now served by 15 hosts, down from 16.
price riseDeepSeek Pro Latest input price rose 4%, $0.190 to $0.198 per million tokens.
price riseDeepSeek Pro Latest output price rose 90%, $4.20 to $8.00 per million tokens.
hostsQwen3 235B A22B Thinking 2507 is now served by 2 hosts, down from 3.
price riseQwen3.6 27B input price rose 41%, $0.320 to $0.450 per million tokens.
price cutQwen3.6 27B output price fell 17%, $3.25 to $2.70 per million tokens.
hostsQwen3.6 27B is now served by 7 hosts, up from 6.
price riseQwen3 30B A3B Instruct 2507 input price rose 99%, $0.048 to $0.100 per million tokens.
price riseQwen3 30B A3B Instruct 2507 output price rose 55%, $0.193 to $0.300 per million tokens.
hostsDeepSeek V3.1 is now served by 5 hosts, down from 7.
price riseGLM 5.2 input price rose 99%, $0.022 to $0.084 per million tokens.
price cutGLM 5.2 output price fell 50%, $16.00 to $8.00 per million tokens.
hostsGLM 5.2 is now served by 34 hosts, up from 32.
price cutClaude Haiku Latest input price fell 90%, $1.00 to $0.100 per million tokens.
price cutClaude Haiku Latest output price fell 90%, $5.00 to $0.500 per million tokens.
contextClaude Haiku Latest context most hosts serve moved from 200K to 1000K.
price cutGLM 5.3 input price fell 22%, $0.050 to $0.039 per million tokens.
hostsGLM 5.3 is now served by 42 hosts, up from 41.
hostsQwen3 VL 235B A22B Thinking is now served by 1 hosts, down from 2.
price riseNemotron 3 Nano 30B A3B input price rose 20%, $0.050 to $0.060 per million tokens.
price riseNemotron 3 Nano 30B A3B output price rose 20%, $0.200 to $0.240 per million tokens.
hostsNemotron 3 Nano 30B A3B is now served by 2 hosts, down from 4.
hostsQwen3 Next 80B A3B Instruct is now served by 4 hosts, down from 5.
hostsGemma 3 12B is now served by 2 hosts, up from 1.
price riseMistral Nemo input price rose 53%, $0.019 to $0.029 per million tokens.
hostsMistral Nemo is now served by 5 hosts, down from 6.
hostsLlama 3 8B Lunaris is now served by 2 hosts, down from 3.
hostsKimi K2.7 Code is now served by 12 hosts, down from 13.
price cutGLM 5.1 input price fell 31%, $1.40 to $0.966 per million tokens.
price cutGLM 5.1 output price fell 31%, $4.40 to $3.04 per million tokens.
price cutDeepSeek V4 Flash 0423 input price fell 75%, $0.030 to $0.007 per million tokens.
hostsDeepSeek V4 Flash 0423 is now served by 17 hosts, up from 16.
price riseKimi Latest input price rose 63%, $0.300 to $0.490 per million tokens.
price cutKimi Latest output price fell 35%, $20.00 to $13.00 per million tokens.
price riseGLM Latest input price rose 4%, $0.030 to $0.031 per million tokens.
price cutGLM Latest output price fell 63%, $12.00 to $4.40 per million tokens.
precisionInferenceNet switched GLM 5.3 from unknown to fp4 — a 4-bit build at the same price.
new hostIo Net started serving GLM 5.3 at $1.20 per million.
precisionInferenceNet switched GLM 5.2 from unknown to fp4 — a 4-bit build at the same price.
new hostNebius started serving GLM 5.2 at $1.40 per million.
precisionPhala switched GLM 5.3 Flash from fp8 to nvfp4.
host leftCloudflare stopped serving GLM 5.3 Flash.
host leftNextBit stopped serving GLM 5.3 Flash.
precisionInferenceNet switched DeepSeek V4.1 Flash from unknown to fp8.
new hostCrusoe started serving DeepSeek V4.1 Flash at $0.290 per million.
new hostOpenInference started serving DeepSeek V4.1 Flash at $0.020 per million.
host leftNextBit stopped serving DeepSeek V4.1 Flash.
host leftNextBit stopped serving DeepSeek V4 Flash 0731.
host leftMorph stopped serving DeepSeek V4 Flash 0731.
precisionMakora switched Kimi K3 from unknown to fp4 — a 4-bit build at the same price.
new hostSail Research started serving DeepSeek V4 Pro 0813 at $0.400 per million.
host leftNextBit stopped serving DeepSeek V4 Pro 0813.
precisionReka switched Qwen3.8 27B from unknown to fp8.
new hostModelRun started serving Qwen3.8 27B at $0.800 per million.
new hostWafer started serving DeepSeek V4 Flash 0423 at $0.035 per million.
host leftDecart stopped serving Kimi K2.6.
host leftNextBit stopped serving DeepSeek V4 Pro 0423.

2026-10-05

202 changes since 2026-09-23

newAion 3.5 appeared, served by 1 host(s) at $3.00 per million input tokens.
newGLM 5.3 Prime appeared, served by 1 host(s) at $2.80 per million input tokens.
newMistral Large 3 2512 appeared, served by 2 host(s) at $0.525 per million input tokens.
newgpt-oss-120b (batch) appeared, served by 23 host(s) at $0.030 per million input tokens.
newPareto 26.10 Preview appeared, served by 1 host(s) at $0.800 per million input tokens.
newPerceptron Mk1.5 appeared, served by 1 host(s) at $0.150 per million input tokens.
newClaude Sonnet 5.5 (batch) appeared, served by 8 host(s) at $1.00 per million input tokens.
newAion 3.5 Mini appeared, served by 1 host(s) at $0.700 per million input tokens.
newGPT-6.1 Sol Pro appeared, served by 6 host(s) at $2.10 per million input tokens.
newEmber-1 appeared, served by 1 host(s) at $3.00 per million input tokens.
newQwen3.8 Max Prime appeared, served by 1 host(s) at $4.00 per million input tokens.
newGPT-6.1 Sol appeared, served by 7 host(s) at $2.20 per million input tokens.
newClaude Sonnet 5.5 appeared, served by 8 host(s) at $2.00 per million input tokens.
withdrawnClaude 3 Haiku is no longer offered by any host we track.
withdrawnR1 Distill Llama 70B is no longer offered by any host we track.
hostsGLM 5.3 Flash is now served by 33 hosts, up from 31.
price riseMiniMax M2 input price rose 18%, $0.255 to $0.300 per million tokens.
price riseMiniMax M2 output price rose 18%, $1.02 to $1.20 per million tokens.
price riseGLM 4.7 input price rose 50%, $0.400 to $0.600 per million tokens.
price riseGLM 4.7 output price rose 26%, $1.75 to $2.20 per million tokens.
hostsGLM 5.3 Flash (batch) is now served by 33 hosts, up from 31.
hostsGemma 4 26B A4B is now served by 13 hosts, up from 12.
price riseQwen3 VL 30B A3B Instruct input price rose 15%, $0.130 to $0.150 per million tokens.
price riseQwen3 VL 30B A3B Instruct output price rose 15%, $0.520 to $0.600 per million tokens.
hostsInkling Small is now served by 1 hosts, down from 2.
price cutMiniMax M2.7 input price fell 30%, $0.300 to $0.210 per million tokens.
price cutMiniMax M2.7 output price fell 30%, $1.20 to $0.840 per million tokens.
hostsMiniMax M2.7 is now served by 7 hosts, down from 9.
price cutDeepSeek V3 input price fell 20%, $0.320 to $0.257 per million tokens.
price riseDeepSeek V3 output price rose 16%, $0.890 to $1.03 per million tokens.
hostsDeepSeek V3.1 Terminus is now served by 3 hosts, down from 4.
contextDeepSeek V3.1 Terminus context most hosts serve moved from 131K to 163K.
price cutDeepSeek V4 Flash Vision Exp input price fell 2%, $0.220 to $0.216 per million tokens.
hostsDeepSeek V4 Flash Vision Exp is now served by 4 hosts, down from 5.
price cutQwen3.5-35B-A3B input price fell 52%, $0.312 to $0.150 per million tokens.
price cutQwen3.5-35B-A3B output price fell 20%, $1.25 to $1.00 per million tokens.
hostsMiMo-V2.5 is now served by 5 hosts, down from 6.
hostsMiMo-V2.5-Pro is now served by 6 hosts, down from 7.
hostsKimi K2.6 is now served by 18 hosts, down from 22.
hostsMistral Small 3.2 24B is now served by 4 hosts, up from 3.
hostsKimi K3 (batch) is now served by 24 hosts, up from 20.
hostsMistral Small 4 is now served by 4 hosts, up from 3.
price cutDeepSeek V4 Pro 0813 input price fell 10%, $0.499 to $0.450 per million tokens.
price riseDeepSeek V4 Pro 0813 output price rose 99%, $1.50 to $5.00 per million tokens.
hostsDeepSeek V4 Pro 0813 is now served by 20 hosts, down from 24.
price riseGrok Latest input price rose 25%, $1.60 to $2.00 per million tokens.
price riseGrok Latest output price rose 25%, $4.80 to $6.00 per million tokens.
hostsGPT-6 Astra Pro (batch) is now served by 6 hosts, up from 5.
price riseDeepSeek V3.2 input price rose 4%, $0.269 to $0.280 per million tokens.
price riseDeepSeek V3.2 output price rose 5%, $0.400 to $0.420 per million tokens.
hostsDeepSeek V3.2 is now served by 13 hosts, down from 15.
price cutDeepSeek Flash Latest input price fell 97%, $0.100 to $0.003 per million tokens.
price riseDeepSeek Flash Latest output price rose 99%, $0.500 to $2.40 per million tokens.
price cutDeepSeek V4 Flash Latest input price fell 60%, $0.038 to $0.015 per million tokens.
price riseDeepSeek V4 Flash Latest output price rose 99%, $0.550 to $1.28 per million tokens.
contextDeepSeek V4 Flash Latest context most hosts serve moved from 1310K to 1048K.
price cutDeepSeek V4 Flash 0731 input price fell 62%, $0.040 to $0.015 per million tokens.
price riseDeepSeek V4 Flash 0731 output price rose 99%, $0.640 to $1.28 per million tokens.
hostsDeepSeek V4 Flash 0731 is now served by 28 hosts, down from 30.
hostsStep 3.7 Flash is now served by 2 hosts, down from 3.
hostsMiMo-V2.6-Pro is now served by 4 hosts, up from 2.
hostsQwen3 32B is now served by 3 hosts, up from 2.
price riseGrok 4.7 input price rose 25%, $1.60 to $2.00 per million tokens.
price riseGrok 4.7 output price rose 25%, $4.80 to $6.00 per million tokens.
hostsGrok 4.7 is now served by 5 hosts, up from 4.
price riseDeepSeek V4.1 Flash input price rose 99%, $0.150 to $0.300 per million tokens.
price riseDeepSeek V4.1 Flash output price rose 99%, $0.600 to $1.20 per million tokens.
hostsDeepSeek V4.1 Flash is now served by 30 hosts, up from 26.
price cutQwen3.8 27B output price fell 15%, $3.00 to $2.55 per million tokens.
hostsQwen3.8 27B is now served by 17 hosts, up from 16.
contextVoxtral Small 24B 2507 context most hosts serve moved from 32K to 32K.
price cutNemotron 3.5 Lightning input price fell 25%, $0.080 to $0.060 per million tokens.
price cutNemotron 3.5 Lightning output price fell 20%, $0.200 to $0.160 per million tokens.
hostsNemotron 3.5 Lightning is now served by 5 hosts, up from 4.
price cutGLM 5.3 (batch) input price fell 38%, $0.720 to $0.450 per million tokens.
price cutGLM 5.3 (batch) output price fell 17%, $2.40 to $2.00 per million tokens.
hostsGLM 5.3 (batch) is now served by 41 hosts, up from 35.
hostsLlama 4 Maverick is now served by 4 hosts, down from 5.
contextLlama 4 Maverick context most hosts serve moved from 1048K to 524K.
price cutGLM Flash Latest input price fell 53%, $0.075 to $0.035 per million tokens.
price riseGLM Flash Latest output price rose 99%, $0.250 to $0.500 per million tokens.
contextGLM Flash Latest context most hosts serve moved from 1310K to 1048K.
hostsMistral Medium 3.5 is now served by 3 hosts, up from 2.
price riseDeepSeek V3 0324 input price rose 16%, $0.250 to $0.290 per million tokens.
price riseDeepSeek V3 0324 output price rose 14%, $1.00 to $1.14 per million tokens.
hostsDeepSeek V3 0324 is now served by 2 hosts, down from 3.
hostsMistral Small 4 (batch) is now served by 4 hosts, up from 3.
hostsDeepSeek V4.1 Flash (batch) is now served by 30 hosts, up from 26.
hostsNano Banana (Gemini 2.5 Flash Image) is now served by 1 hosts, down from 4.
price cutNemotron 3 Ultra input price fell 17%, $0.600 to $0.500 per million tokens.
price cutNemotron 3 Ultra output price fell 8%, $2.40 to $2.20 per million tokens.
hostsHy4 preview is now served by 4 hosts, up from 2.
hostsMiMo-V2.6-Flash is now served by 7 hosts, up from 2.
price cutInkling input price fell 5%, $1.00 to $0.950 per million tokens.
hostsInkling is now served by 2 hosts, down from 4.
price cutgpt-oss-120b input price fell 75%, $0.150 to $0.037 per million tokens.
price cutgpt-oss-120b output price fell 72%, $0.600 to $0.170 per million tokens.
hostsgpt-oss-120b is now served by 23 hosts, down from 24.
hostsGPT-6 Astra is now served by 7 hosts, up from 6.
hostsGPT-6 Astra Pro is now served by 6 hosts, up from 5.
price cutKimi K3 input price fell 78%, $3.00 to $0.670 per million tokens.
price cutKimi K3 output price fell 7%, $15.00 to $14.00 per million tokens.
hostsKimi K3 is now served by 24 hosts, up from 20.
hostsDeepSeek V3.2 Exp is now served by 1 hosts, down from 3.
hostsGPT-6 Astra (batch) is now served by 7 hosts, up from 6.
price riseMiniMax M1 input price rose 38%, $0.400 to $0.550 per million tokens.
price riseMuse Glimmer 30B input price rose 17%, $0.300 to $0.350 per million tokens.
price riseMuse Glimmer 30B output price rose 25%, $1.20 to $1.50 per million tokens.
hostsMuse Glimmer 30B is now served by 3 hosts, down from 4.
hostsQwen3.5-122B-A10B is now served by 4 hosts, down from 5.
hostsGrok 4.6 is now served by 6 hosts, up from 5.
price cutDeepSeek V4 Pro 0423 input price fell 78%, $0.955 to $0.209 per million tokens.
price cutDeepSeek V4 Pro 0423 output price fell 78%, $1.91 to $0.418 per million tokens.
price cutDeepSeek Pro Latest input price fell 52%, $0.400 to $0.190 per million tokens.
price cutDeepSeek Pro Latest output price fell 2%, $4.30 to $4.20 per million tokens.
price cutLing 3.0 Flash Fin input price fell 30%, $0.060 to $0.042 per million tokens.
price cutLing 3.0 Flash Fin output price fell 32%, $0.180 to $0.123 per million tokens.
hostsLing 3.0 Flash Fin is now served by 2 hosts, up from 1.
price riseQwen3.6 27B output price rose 20%, $2.70 to $3.25 per million tokens.
hostsMistral Medium 3.5 (batch) is now served by 3 hosts, up from 2.
hostsDeepSeek V3.1 is now served by 7 hosts, down from 8.
contextDeepSeek V3.1 context most hosts serve moved from 131K to 163K.
price cutGLM 5.2 input price fell 97%, $0.650 to $0.022 per million tokens.
price riseGLM 5.2 output price rose 99%, $2.04 to $16.00 per million tokens.
hostsGLM 5.2 is now served by 32 hosts, up from 29.
price cutLing 3.0 Flash VL input price fell 65%, $0.060 to $0.021 per million tokens.
price cutLing 3.0 Flash VL output price fell 66%, $0.180 to $0.062 per million tokens.
hostsLing 3.0 Flash VL is now served by 2 hosts, up from 1.
contextLing 3.0 Flash VL context most hosts serve moved from 131K to 262K.
price cutGLM 5.3 input price fell 94%, $0.840 to $0.050 per million tokens.
price riseGLM 5.3 output price rose 99%, $2.64 to $7.00 per million tokens.
hostsGLM 5.3 is now served by 41 hosts, up from 35.
price riseQwen3 Next 80B A3B Instruct input price rose 11%, $0.090 to $0.100 per million tokens.
hostsMiniMax M2.5 is now served by 9 hosts, up from 8.
contextMiniMax M2.5 context most hosts serve moved from 204K to 196K.
hostsMistral Nemo is now served by 6 hosts, up from 4.
price cutKimi K2.7 Code input price fell 5%, $0.706 to $0.671 per million tokens.
hostsKimi K2.7 Code is now served by 13 hosts, down from 15.
price riseGLM 5.1 input price rose 45%, $0.966 to $1.40 per million tokens.
price riseGLM 5.1 output price rose 45%, $3.04 to $4.40 per million tokens.
hostsGLM 5.1 is now served by 13 hosts, down from 14.
price cutDeepSeek V4 Flash 0423 input price fell 66%, $0.089 to $0.030 per million tokens.
price riseDeepSeek V4 Flash 0423 output price rose 99%, $0.177 to $1.28 per million tokens.
hostsDeepSeek V4 Flash 0423 is now served by 16 hosts, up from 15.
price cutKimi Latest input price fell 80%, $1.50 to $0.300 per million tokens.
price riseKimi Latest output price rose 86%, $10.76 to $20.00 per million tokens.
price cutGLM Latest input price fell 95%, $0.562 to $0.030 per million tokens.
price riseGLM Latest output price rose 99%, $2.50 to $12.00 per million tokens.
contextGLM Latest context most hosts serve moved from 1310K to 1048K.
precisionInferenceNet switched GLM 5.3 from fp4 to unknown.
precisionMorph switched GLM 5.3 from unknown to fp8.
precisionReka switched GLM 5.3 from fp8 to unknown.
new hostNebius started serving GLM 5.3 at $1.40 per million.
new hostRelace started serving GLM 5.3 at $0.030 per million.
host leftIo Net stopped serving GLM 5.3.
precisionSail Research switched GLM 5.3 Flash from fp8 to fp4 — a 4-bit build at the same price.
precisionMorph switched GLM 5.3 Flash from unknown to fp8.
precisionModal switched GLM 5.3 Flash from fp8 to nvfp4.
precisionReka switched GLM 5.3 Flash from fp8 to unknown.
precisionParasail switched GLM 5.3 Flash from fp8 to fp4 — a 4-bit build at the same price.
new hostDecart started serving GLM 5.3 Flash at $0.090 per million.
new hostDekaLLM started serving GLM 5.3 Flash at $0.100 per million.
host leftIo Net stopped serving GLM 5.3 Flash.
new hostMistral started serving GLM 5.2 at $1.54 per million.
new hostInferenceNet started serving GLM 5.2 at $0.028 per million.
new hostRelace started serving GLM 5.2 at $0.022 per million.
new hostMorph started serving GLM 5.2 at $0.170 per million.
host leftFireworks stopped serving GLM 5.2.
precisionRelace switched DeepSeek V4.1 Flash from fp4 to unknown.
precisionMorph switched DeepSeek V4.1 Flash from unknown to fp8.
precisionMakora switched DeepSeek V4.1 Flash from unknown to fp8.
new hostInferenceNet started serving DeepSeek V4.1 Flash at $0.070 per million.
new hostIonstream started serving DeepSeek V4.1 Flash at $0.119 per million.
new hostDecart started serving DeepSeek V4.1 Flash at $0.090 per million.
new hostNextBit started serving DeepSeek V4.1 Flash at $0.210 per million.
new hostBaidu started serving DeepSeek V4.1 Flash at $0.300 per million.
host leftKrea stopped serving DeepSeek V4.1 Flash.
host leftOpenInference stopped serving DeepSeek V4.1 Flash.
precisionOpenInference switched DeepSeek V4 Flash 0731 from fp8 to fp4 — a 4-bit build at the same price.
precisionReka switched DeepSeek V4 Flash 0731 from fp4 to unknown.
precisionMorph switched DeepSeek V4 Flash 0731 from unknown to bf16.
host leftMakora stopped serving DeepSeek V4 Flash 0731.
host leftFireworks stopped serving DeepSeek V4 Flash 0731.
host leftNebius stopped serving DeepSeek V4 Flash 0731.
precisionMorph switched Kimi K3 from unknown to fp8.
precisionDeepInfra switched Kimi K3 from bf16 to mxfp4.
new hostAmazon Bedrock started serving Kimi K3 at $3.00 per million.
new hostDecart started serving Kimi K3 at $1.80 per million.
new hostAkashML started serving Kimi K3 at $3.00 per million.
precisionDeepInfra switched gpt-oss-120b from fp8 to bf16.
new hostRelace started serving DeepSeek V4 Pro 0813 at $0.190 per million.
host leftBaseTen stopped serving DeepSeek V4 Pro 0813.
host leftFireworks stopped serving DeepSeek V4 Pro 0813.
host leftSail Research stopped serving DeepSeek V4 Pro 0813.
precisionReka switched Qwen3.8 27B from fp8 to unknown.
new hostCerebras started serving Qwen3.8 27B at $0.990 per million.
host leftBaseTen stopped serving Kimi K2.6.
host leftFireworks stopped serving Kimi K2.6.
host leftSail Research stopped serving Kimi K2.6.
new hostRelace started serving DeepSeek V4 Pro 0423 at $0.207 per million.
new hostReka started serving DeepSeek V4 Pro 0423 at $0.812 per million.
host leftBaseTen stopped serving DeepSeek V4 Pro 0423.

2026-09-23

218 changes since 2026-09-09

newGPT-6 Sol appeared, served by 7 host(s) at $2.20 per million input tokens.
newGPT-6 Luna appeared, served by 7 host(s) at $0.110 per million input tokens.
newGPT-6 Luna (batch) appeared, served by 7 host(s) at $0.050 per million input tokens.
newMiMo-V2.6-Pro-UltraSpeed appeared, served by 1 host(s) at $4.35 per million input tokens.
newGPT-6 Sol (batch) appeared, served by 7 host(s) at $1.00 per million input tokens.
newSchematron V2 Turbo appeared, served by 1 host(s) at $0.030 per million input tokens.
newMistral Medium 3.1 (batch) appeared, served by 3 host(s) at $0.200 per million input tokens.
newDeepSeek Pro Latest appeared, served by 0 host(s) at $0.400 per million input tokens.
newGPT-6 Luna Pro (batch) appeared, served by 6 host(s) at $0.050 per million input tokens.
newGrok 4.7 appeared, served by 4 host(s) at $2.40 per million input tokens.
newTernary Bonsai 2 27B appeared, served by 1 host(s) at $0.075 per million input tokens.
newDeepSeek V4.1 Flash appeared, served by 26 host(s) at $0.280 per million input tokens.
newGPT Astra Latest appeared, served by 0 host(s) at $10.00 per million input tokens.
newSolar Mini 4 appeared, served by 2 host(s) at $0.050 per million input tokens.
newGPT Luna Latest appeared, served by 0 host(s) at $0.100 per million input tokens.
newGPT Terra Latest appeared, served by 0 host(s) at $2.00 per million input tokens.
newClaude Opus 5.5 appeared, served by 11 host(s) at $4.40 per million input tokens.
newQwen3.8 Omni Flash appeared, served by 1 host(s) at $0.150 per million input tokens.
newGPT-6 Sol Pro (batch) appeared, served by 6 host(s) at $1.00 per million input tokens.
newCommand A+ appeared, served by 1 host(s) at $0.300 per million input tokens.
newCodestral 2508 (batch) appeared, served by 2 host(s) at $0.150 per million input tokens.
newMistral Small 4 (batch) appeared, served by 3 host(s) at $0.075 per million input tokens.
newMistral Large 3 2512 (batch) appeared, served by 0 host(s) at $0.250 per million input tokens.
newDeepSeek Flash Latest appeared, served by 0 host(s) at $0.100 per million input tokens.
newClaude Opus 5.5 (batch) appeared, served by 11 host(s) at $2.00 per million input tokens.
newNex-N2.5-Mini appeared, served by 1 host(s) at $0.025 per million input tokens.
newGPT-6 Luna Pro appeared, served by 6 host(s) at $0.105 per million input tokens.
newDeepSeek V4.1 Flash (batch) appeared, served by 26 host(s) at $0.112 per million input tokens.
newSchematron V2 Small appeared, served by 1 host(s) at $0.050 per million input tokens.
newMinistral 3 8B 2512 (batch) appeared, served by 3 host(s) at $0.075 per million input tokens.
newLing 3.0 Flash VL appeared, served by 1 host(s) at $0.060 per million input tokens.
newFugu Ultra v2 appeared, served by 1 host(s) at $5.00 per million input tokens.
newMiMo-V2.6-Flash appeared, served by 2 host(s) at $0.140 per million input tokens.
newNex-N2.5-Pro appeared, served by 1 host(s) at $0.075 per million input tokens.
newGPT-6 Sol Pro appeared, served by 6 host(s) at $2.10 per million input tokens.
newMiMo-V2.6-Pro appeared, served by 2 host(s) at $0.435 per million input tokens.
newFugu Max appeared, served by 1 host(s) at $2.00 per million input tokens.
newGPT Sol Latest appeared, served by 0 host(s) at $2.00 per million input tokens.
newGLM 5.3 FlashX appeared, served by 1 host(s) at $0.370 per million input tokens.
newPareto appeared, served by 1 host(s) at $2.50 per million input tokens.
withdrawnOpenAI GPT Latest is no longer offered by any host we track.
withdrawnInkling (batch) is no longer offered by any host we track.
withdrawnQwen3.8 2.4T A95B (batch) is no longer offered by any host we track.
withdrawnKAT-Coder-Pro V2 is no longer offered by any host we track.
withdrawnHermes 4 70B is no longer offered by any host we track.
withdrawnGemma 4 31B (batch) is no longer offered by any host we track.
withdrawnInkling Small (batch) is no longer offered by any host we track.
withdrawnGemini 2.5 Pro Preview 05-06 is no longer offered by any host we track.
withdrawnGPT-4 Turbo Preview is no longer offered by any host we track.
withdrawnQwen3.5-9B (batch) is no longer offered by any host we track.
withdrawnClaude Opus 4 is no longer offered by any host we track.
withdrawnDeepSeek V4 Pro 0813 (batch) is no longer offered by any host we track.
withdrawnMiniMax M3 (batch) is no longer offered by any host we track.
withdrawnGLM 5.2 (batch) is no longer offered by any host we track.
withdrawnDeepSeek V4 Flash Vision Exp (batch) is no longer offered by any host we track.
withdrawnMuse Glimmer 30B (batch) is no longer offered by any host we track.
withdrawnDeepSeek V4 Flash 0731 (batch) is no longer offered by any host we track.
price riseGLM 5.3 Flash input price rose 99%, $0.075 to $0.150 per million tokens.
price riseGLM 5.3 Flash output price rose 99%, $0.250 to $0.500 per million tokens.
hostsGLM 5.3 Flash is now served by 31 hosts, up from 25.
price cutGLM 5.3 Flash (batch) input price fell 20%, $0.075 to $0.060 per million tokens.
price cutGLM 5.3 Flash (batch) output price fell 20%, $0.250 to $0.200 per million tokens.
hostsGLM 5.3 Flash (batch) is now served by 31 hosts, up from 25.
hostsMiniMax M3 is now served by 13 hosts, up from 12.
price riseGemma 4 26B A4B input price rose 29%, $0.070 to $0.090 per million tokens.
price cutGemma 4 26B A4B output price fell 12%, $0.340 to $0.300 per million tokens.
hostsGemma 4 26B A4B is now served by 12 hosts, up from 9.
price cutQwen3 VL 30B A3B Instruct input price fell 13%, $0.150 to $0.130 per million tokens.
price cutQwen3 VL 30B A3B Instruct output price fell 13%, $0.600 to $0.520 per million tokens.
hostsInkling Small is now served by 2 hosts, down from 3.
lost best valueGemini 3.6 Flash (batch) lost best value; something cheaper now matches its score.
price riseSolar Pro 4 input price rose 99%, $0.030 to $0.090 per million tokens.
price riseSolar Pro 4 output price rose 99%, $0.120 to $0.360 per million tokens.
lost best valueSolar Pro 4 lost best value; something cheaper now matches its score.
hostsClaude Opus 4.1 is now served by 1 hosts, down from 2.
price cutQwen3 235B A22B Instruct 2507 input price fell 60%, $0.220 to $0.087 per million tokens.
price cutQwen3 235B A22B Instruct 2507 output price fell 60%, $0.880 to $0.350 per million tokens.
hostsQwen3 235B A22B Instruct 2507 is now served by 10 hosts, down from 11.
hostsMiniMax M2.7 is now served by 9 hosts, down from 10.
hostsKimi K2.5 is now served by 5 hosts, down from 6.
hostsDeepSeek V4 Flash Vision Exp is now served by 5 hosts, down from 6.
hostsGLM 4.7 Flash is now served by 3 hosts, down from 4.
hostsMiMo-V2.5 is now served by 6 hosts, up from 5.
lost best valueMiMo-V2.5 lost best value; something cheaper now matches its score.
price cutQwen3 14B input price fell 47%, $0.227 to $0.120 per million tokens.
price cutQwen3 14B output price fell 74%, $0.910 to $0.240 per million tokens.
hostsKimi K2.6 is now served by 22 hosts, up from 21.
price riseMistral Small 3.2 24B input price rose 25%, $0.075 to $0.094 per million tokens.
price riseMistral Small 3.2 24B output price rose 25%, $0.200 to $0.250 per million tokens.
hostsMistral Small 3.2 24B is now served by 3 hosts, up from 2.
price cutKimi K3 (batch) input price fell 24%, $3.00 to $2.28 per million tokens.
price cutKimi K3 (batch) output price fell 24%, $15.00 to $11.40 per million tokens.
hostsKimi K3 (batch) is now served by 20 hosts, up from 19.
hostsMistral Small 4 is now served by 3 hosts, down from 4.
price cutDeepSeek V4 Pro 0813 input price fell 14%, $0.579 to $0.499 per million tokens.
price cutDeepSeek V4 Pro 0813 output price fell 14%, $1.74 to $1.50 per million tokens.
hostsDeepSeek V4 Pro 0813 is now served by 24 hosts, up from 20.
price cutGrok Latest input price fell 20%, $2.00 to $1.60 per million tokens.
price cutGrok Latest output price fell 20%, $6.00 to $4.80 per million tokens.
price cutNemotron 3 Super input price fell 6%, $0.085 to $0.080 per million tokens.
price riseNemotron 3 Super output price rose 12%, $0.400 to $0.450 per million tokens.
hostsNemotron 3 Super is now served by 2 hosts, up from 1.
hostsClaude Opus 5 (batch) is now served by 11 hosts, up from 10.
hostsLlama 3.3 70B Instruct is now served by 11 hosts, down from 13.
price cutDeepSeek V4 Flash Latest input price fell 24%, $0.050 to $0.038 per million tokens.
price riseDeepSeek V4 Flash Latest output price rose 99%, $0.160 to $0.550 per million tokens.
price cutDeepSeek V4 Flash 0731 input price fell 38%, $0.065 to $0.040 per million tokens.
price riseDeepSeek V4 Flash 0731 output price rose 99%, $0.180 to $0.640 per million tokens.
hostsDeepSeek V4 Flash 0731 is now served by 30 hosts, up from 29.
hostsQwen3.8 27B is now served by 16 hosts, up from 13.
hostsNemotron 3.5 Lightning is now served by 4 hosts, up from 2.
hostsClaude Sonnet 5 (batch) is now served by 10 hosts, up from 9.
hostsGLM 5 is now served by 8 hosts, down from 9.
price cutgpt-oss-20b (batch) input price fell 52%, $0.050 to $0.024 per million tokens.
price cutgpt-oss-20b (batch) output price fell 44%, $0.200 to $0.112 per million tokens.
hostsgpt-oss-20b (batch) is now served by 12 hosts, down from 13.
best valuegpt-oss-20b (batch) became best value — nothing cheaper now matches it.
price riseGLM 5.3 (batch) input price rose 3%, $0.700 to $0.720 per million tokens.
price riseGLM 5.3 (batch) output price rose 9%, $2.20 to $2.40 per million tokens.
hostsGLM 5.3 (batch) is now served by 35 hosts, up from 28.
price cutLlama 4 Maverick input price fell 6%, $0.200 to $0.188 per million tokens.
price cutLlama 4 Maverick output price fell 6%, $0.696 to $0.652 per million tokens.
hostsMistral Medium 3.5 is now served by 2 hosts, down from 3.
price cutDeepSeek V3 0324 input price fell 14%, $0.290 to $0.250 per million tokens.
price cutDeepSeek V3 0324 output price fell 12%, $1.14 to $1.00 per million tokens.
hostsDeepSeek V3 0324 is now served by 3 hosts, down from 4.
hostsUnslopNemo 12B is now served by 1 hosts, down from 2.
hostsClaude Sonnet 5 is now served by 10 hosts, up from 9.
price riseQwen3.6 35B A3B input price rose 50%, $0.100 to $0.150 per million tokens.
price riseQwen3.6 35B A3B output price rose 11%, $0.900 to $1.00 per million tokens.
hostsQwen3.6 35B A3B is now served by 9 hosts, down from 10.
lost best valueGemini 3.7 Flash (batch) lost best value; something cheaper now matches its score.
price cutNemotron 3 Ultra input price fell 4%, $0.625 to $0.600 per million tokens.
price cutNemotron 3 Ultra output price fell 23%, $3.12 to $2.40 per million tokens.
hostsHy4 preview is now served by 2 hosts, up from 1.
hostsClaude Opus 5 is now served by 11 hosts, up from 10.
price risegpt-oss-120b input price rose 99%, $0.037 to $0.150 per million tokens.
price risegpt-oss-120b output price rose 99%, $0.170 to $0.600 per million tokens.
hostsgpt-oss-120b is now served by 24 hosts, up from 22.
price cutgpt-oss-20b input price fell 40%, $0.030 to $0.018 per million tokens.
price cutgpt-oss-20b output price fell 31%, $0.130 to $0.090 per million tokens.
hostsgpt-oss-20b is now served by 12 hosts, down from 13.
hostsGPT-6 Astra is now served by 6 hosts, up from 5.
hostsClaude Sonnet 4 is now served by 2 hosts, down from 5.
contextClaude Sonnet 4 context most hosts serve moved from 1000K to 200K.
hostsKimi K3 is now served by 20 hosts, up from 19.
hostsGPT-6 Astra (batch) is now served by 6 hosts, up from 5.
price cutMiniMax M1 input price fell 27%, $0.550 to $0.400 per million tokens.
price riseMuse Glimmer 30B output price rose 9%, $1.10 to $1.20 per million tokens.
price cutQwen3.5-122B-A10B input price fell 10%, $0.290 to $0.260 per million tokens.
price cutQwen3.5-122B-A10B output price fell 13%, $2.40 to $2.08 per million tokens.
price cutClaude Opus Latest input price fell 20%, $5.00 to $4.00 per million tokens.
price cutClaude Opus Latest output price fell 20%, $25.00 to $20.00 per million tokens.
price riseQwen3.6 27B input price rose 7%, $0.300 to $0.320 per million tokens.
price riseQwen3.6 27B output price rose 35%, $2.00 to $2.70 per million tokens.
hostsMistral Medium 3.5 (batch) is now served by 2 hosts, down from 3.
hostsGLM 4.6 is now served by 4 hosts, down from 5.
hostsQwen3 30B A3B Instruct 2507 is now served by 5 hosts, up from 4.
price riseMythoMax 13B input price rose 33%, $0.060 to $0.080 per million tokens.
price riseMythoMax 13B output price rose 83%, $0.060 to $0.110 per million tokens.
hostsMythoMax 13B is now served by 3 hosts, down from 4.
price cutGLM 5.2 input price fell 33%, $0.966 to $0.650 per million tokens.
price cutGLM 5.2 output price fell 33%, $3.04 to $2.04 per million tokens.
hostsGLM 5.2 is now served by 29 hosts, down from 33.
price cutGLM 5.3 input price fell 40%, $1.40 to $0.840 per million tokens.
price cutGLM 5.3 output price fell 40%, $4.40 to $2.64 per million tokens.
hostsGLM 5.3 is now served by 35 hosts, up from 28.
hostsClaude Opus 4.1 (batch) is now served by 1 hosts, down from 2.
hostsGemma 4 31B is now served by 13 hosts, down from 15.
hostsMiniMax M2.5 is now served by 8 hosts, down from 9.
contextMiniMax M2.5 context most hosts serve moved from 196K to 204K.
lost best valueGemini 3.8 Flash (batch) lost best value; something cheaper now matches its score.
price cutKimi K2.7 Code output price fell 6%, $3.50 to $3.30 per million tokens.
hostsGLM 5.1 is now served by 14 hosts, down from 15.
price cutKimi Latest input price fell 38%, $2.40 to $1.50 per million tokens.
price cutKimi Latest output price fell 10%, $12.00 to $10.76 per million tokens.
price cutGLM Latest input price fell 49%, $1.11 to $0.562 per million tokens.
price cutGLM Latest output price fell 29%, $3.50 to $2.50 per million tokens.
precisionMorph switched GLM 5.3 Flash from fp8 to unknown.
precisionRelace switched GLM 5.3 Flash from fp4 to unknown.
precisionCoreWeave switched GLM 5.3 Flash from fp8 to nvfp4.
new hostAtlasCloud started serving GLM 5.3 Flash at $0.150 per million.
new hostInferenceNet started serving GLM 5.3 Flash at $0.090 per million.
new hostNear AI started serving GLM 5.3 Flash at $0.150 per million.
new hostReka started serving GLM 5.3 Flash at $0.150 per million.
new hostInceptron started serving GLM 5.3 Flash at $0.150 per million.
new hostOpenInference started serving GLM 5.3 Flash at $0.100 per million.
host leftMakora stopped serving GLM 5.3 Flash.
precisionIo Net switched GLM 5.3 from fp4 to fp8.
precisionMorph switched GLM 5.3 from fp4 to unknown.
new hostMistral started serving GLM 5.3 at $1.40 per million.
new hostInferenceNet started serving GLM 5.3 at $0.900 per million.
new hostPrimeIntellect started serving GLM 5.3 at $1.40 per million.
new hostAlibaba started serving GLM 5.3 at $1.19 per million.
new hostBaidu started serving GLM 5.3 at $1.40 per million.
host leftStreamLake stopped serving GLM 5.3.
precisionMorph switched DeepSeek V4 Flash 0731 from bf16 to unknown.
new hostCohere started serving DeepSeek V4 Flash 0731 at $0.140 per million.
new hostNebius started serving DeepSeek V4 Flash 0731 at $0.140 per million.
host leftDeepSeek stopped serving DeepSeek V4 Flash 0731.
new hostWafer started serving GLM 5.2 at $1.40 per million.
host leftMistral stopped serving GLM 5.2.
host leftAmbient stopped serving GLM 5.2.
host leftCrusoe stopped serving GLM 5.2.
precisionIonstream switched DeepSeek V4 Pro 0813 from fp4 to unknown.
new hostAtlasCloud started serving DeepSeek V4 Pro 0813 at $1.32 per million.
new hostVenice started serving DeepSeek V4 Pro 0813 at $1.65 per million.
new hostWafer started serving DeepSeek V4 Pro 0813 at $0.500 per million.
new hostDekaLLM started serving gpt-oss-120b at $0.030 per million.
new hostCrusoe started serving gpt-oss-120b at $0.050 per million.
new hostSail Research started serving Kimi K2.6 at $1.00 per million.
precisionMorph switched Kimi K3 from fp4 to unknown.
new hostInferenceNet started serving Kimi K3 at $1.95 per million.
new hostDekaLLM started serving gpt-oss-20b at $0.029 per million.
host leftTogether stopped serving gpt-oss-20b.
host leftPhala stopped serving gpt-oss-20b.
new hostCloudflare started serving DeepSeek V4 Pro 0423 at $1.15 per million.
host leftFireworks stopped serving DeepSeek V4 Pro 0423.

2026-09-09

82 changes since 2026-09-07

newGLM 5.3 (batch) appeared, served by 28 host(s) at $0.700 per million input tokens.
newGLM 5.2 (batch) appeared, served by 33 host(s) at $0.700 per million input tokens.
newDeepSeek V4 Flash Vision Exp (batch) appeared, served by 6 host(s) at $0.110 per million input tokens.
newMercury 2.5 appeared, served by 1 host(s) at $0.040 per million input tokens.
withdrawnNex-N2-Pro is no longer offered by any host we track.
withdrawnNex-N2-Mini is no longer offered by any host we track.
withdrawnMercury 2.5 Preview is no longer offered by any host we track.
hostsGLM 5.3 Flash is now served by 25 hosts, up from 23.
price cutGLM 5.3 Flash (batch) input price fell 50%, $0.150 to $0.075 per million tokens.
price cutGLM 5.3 Flash (batch) output price fell 50%, $0.500 to $0.250 per million tokens.
hostsGLM 5.3 Flash (batch) is now served by 25 hosts, up from 23.
contextMiniMax M3 context most hosts serve moved from 524K to 1048K.
price riseQwen3 235B A22B Instruct 2507 input price rose 99%, $0.090 to $0.220 per million tokens.
price riseQwen3 235B A22B Instruct 2507 output price rose 60%, $0.550 to $0.880 per million tokens.
price riseQwen3.5 397B A17B input price rose 41%, $0.390 to $0.550 per million tokens.
price riseQwen3.5 397B A17B output price rose 50%, $2.34 to $3.50 per million tokens.
hostsKimi K2.5 is now served by 6 hosts, down from 8.
best valueMiMo-V2.5 became best value — nothing cheaper now matches it.
price riseQwen3 14B input price rose 90%, $0.120 to $0.227 per million tokens.
price riseQwen3 14B output price rose 99%, $0.240 to $0.910 per million tokens.
hostsKimi K3 (batch) is now served by 19 hosts, up from 18.
lost best valueKimi K3 (batch) lost best value; something cheaper now matches its score.
price cutGranite 4.2 8B input price fell 40%, $0.100 to $0.060 per million tokens.
price riseGranite 4.2 8B output price rose 67%, $0.150 to $0.250 per million tokens.
hostsDeepSeek V4 Pro 0813 is now served by 20 hosts, up from 19.
hostsNemotron 3 Super is now served by 1 hosts, down from 2.
price riseHy3 input price rose 60%, $0.083 to $0.132 per million tokens.
price riseHy3 output price rose 60%, $0.330 to $0.528 per million tokens.
price riseDeepSeek V4 Flash Latest input price rose 11%, $0.045 to $0.050 per million tokens.
price riseDeepSeek V4 Flash Latest output price rose 78%, $0.090 to $0.160 per million tokens.
price riseDeepSeek V4 Flash 0731 input price rose 30%, $0.050 to $0.065 per million tokens.
price riseDeepSeek V4 Flash 0731 output price rose 80%, $0.100 to $0.180 per million tokens.
lost best valueDeepSeek V4 Flash 0731 lost best value; something cheaper now matches its score.
hostsGLM 5 is now served by 9 hosts, down from 10.
hostsgpt-oss-20b (batch) is now served by 13 hosts, down from 14.
price riseDeepSeek V3 0324 input price rose 16%, $0.250 to $0.290 per million tokens.
price riseDeepSeek V3 0324 output price rose 14%, $1.00 to $1.14 per million tokens.
hostsgpt-oss-20b is now served by 13 hosts, down from 14.
price cutDeepSeek V4 Pro 0813 (batch) input price fell 50%, $1.32 to $0.660 per million tokens.
price cutDeepSeek V4 Pro 0813 (batch) output price fell 50%, $3.96 to $1.98 per million tokens.
hostsDeepSeek V4 Pro 0813 (batch) is now served by 20 hosts, up from 19.
contextMiniMax M3 (batch) context most hosts serve moved from 524K to 1048K.
hostsKimi K3 is now served by 19 hosts, up from 18.
lost best valueKimi K3 lost best value; something cheaper now matches its score.
hostsQwen3.5-9B is now served by 6 hosts, up from 5.
price riseDeepSeek V4 Pro 0423 input price rose 50%, $0.637 to $0.955 per million tokens.
price riseDeepSeek V4 Pro 0423 output price rose 50%, $1.27 to $1.91 per million tokens.
hostsDeepSeek V4 Pro 0423 is now served by 16 hosts, down from 17.
price cutGLM 4.6 input price fell 22%, $0.550 to $0.430 per million tokens.
price cutGLM 4.6 output price fell 20%, $2.20 to $1.75 per million tokens.
hostsQwen3.5-9B (batch) is now served by 6 hosts, up from 5.
price cutDeepSeek V4 Flash 0731 (batch) input price fell 21%, $0.140 to $0.110 per million tokens.
price riseDeepSeek V4 Flash 0731 (batch) output price rose 18%, $0.280 to $0.330 per million tokens.
best valueDeepSeek V4 Flash 0731 (batch) became best value — nothing cheaper now matches it.
price cutDeepSeek V3.1 input price fell 55%, $0.550 to $0.250 per million tokens.
price cutDeepSeek V3.1 output price fell 42%, $1.65 to $0.950 per million tokens.
hostsGLM 5.3 is now served by 28 hosts, up from 27.
price cutQwen3 Next 80B A3B Instruct input price fell 10%, $0.100 to $0.090 per million tokens.
price riseKimi K2.7 Code input price rose 8%, $0.660 to $0.710 per million tokens.
price riseKimi K2.7 Code output price rose 3%, $3.40 to $3.50 per million tokens.
hostsKimi K2.7 Code is now served by 15 hosts, down from 16.
hostsGLM 5.1 is now served by 15 hosts, down from 16.
price riseDeepSeek V4 Flash 0423 input price rose 11%, $0.080 to $0.089 per million tokens.
price riseDeepSeek V4 Flash 0423 output price rose 11%, $0.159 to $0.177 per million tokens.
price cutMoonshotAI Kimi Latest input price fell 6%, $2.55 to $2.40 per million tokens.
price cutMoonshotAI Kimi Latest output price fell 6%, $12.75 to $12.00 per million tokens.
price cutGLM Latest input price fell 5%, $1.17 to $1.11 per million tokens.
price cutGLM Latest output price fell 12%, $3.96 to $3.50 per million tokens.
price cutMuse Glimmer 30B (batch) input price fell 50%, $0.350 to $0.175 per million tokens.
price cutMuse Glimmer 30B (batch) output price fell 50%, $1.50 to $0.750 per million tokens.
precisionIo Net switched GLM 5.3 from fp8 to fp4 — a 4-bit build at the same price.
precisionCrusoe switched GLM 5.3 from fp8 to fp4 — a 4-bit build at the same price.
new hostStreamLake started serving GLM 5.3 at $1.11 per million.
precisionBaidu switched GLM 5.2 from fp8 to fp4 — a 4-bit build at the same price.
precisionStreamLake switched GLM 5.3 Flash from unknown to fp8.
new hostCrusoe started serving GLM 5.3 Flash at $0.150 per million.
new hostRelace started serving GLM 5.3 Flash at $0.090 per million.
new hostPhala started serving GLM 5.3 Flash at $0.150 per million.
host leftReka stopped serving GLM 5.3 Flash.
new hostIonstream started serving DeepSeek V4 Pro 0813 at $1.10 per million.
new hostRelace started serving Kimi K3 at $2.40 per million.
host leftIonstream stopped serving DeepSeek V4 Pro 0423.

2026-09-07

11 changes since 2026-09-06

price cutQwen3.5 397B A17B input price fell 29%, $0.550 to $0.390 per million tokens.
price cutQwen3.5 397B A17B output price fell 33%, $3.50 to $2.34 per million tokens.
price cutDeepSeek V4 Pro 0813 input price fell 48%, $1.12 to $0.579 per million tokens.
price cutDeepSeek V4 Pro 0813 output price fell 48%, $3.36 to $1.74 per million tokens.
price cutHy3 input price fell 37%, $0.132 to $0.083 per million tokens.
price cutHy3 output price fell 37%, $0.528 to $0.330 per million tokens.
price cutDeepSeek V4 Pro 0423 input price fell 9%, $0.698 to $0.637 per million tokens.
price cutDeepSeek V4 Pro 0423 output price fell 9%, $1.40 to $1.27 per million tokens.
hostsGLM 5.2 is now served by 33 hosts, up from 32.
hostsGLM 5.3 is now served by 27 hosts, up from 26.
new hostCrusoe started serving GLM 5.3 at $1.40 per million.

2026-09-06

25 changes since 2026-09-04

newGPT-6 Astra Pro (batch) appeared, served by 5 host(s) at $5.00 per million input tokens.
newGPT-6 Astra appeared, served by 5 host(s) at $10.00 per million input tokens.
newQwen3.8 Max (0902) appeared, served by 1 host(s) at $2.00 per million input tokens.
newGPT-6 Astra (batch) appeared, served by 5 host(s) at $5.00 per million input tokens.
newGPT-6 Astra Pro appeared, served by 5 host(s) at $10.00 per million input tokens.
withdrawnGranite 4.1 8B is no longer offered by any host we track.
withdrawnQwen3.8 Max is no longer offered by any host we track.
price riseQwen3 235B A22B Instruct 2507 input price rose 3%, $0.087 to $0.090 per million tokens.
price riseQwen3 235B A22B Instruct 2507 output price rose 57%, $0.350 to $0.550 per million tokens.
price cutDeepSeek V4 Flash Vision Exp input price fell 50%, $0.440 to $0.220 per million tokens.
price cutDeepSeek V4 Flash Vision Exp output price fell 50%, $1.32 to $0.660 per million tokens.
price riseQwen3.5-35B-A3B input price rose 25%, $0.250 to $0.312 per million tokens.
price cutDeepSeek V4 Flash Latest input price fell 10%, $0.050 to $0.045 per million tokens.
price cutDeepSeek V4 Flash Latest output price fell 44%, $0.160 to $0.090 per million tokens.
price cutDeepSeek V4 Flash 0731 input price fell 23%, $0.065 to $0.050 per million tokens.
price cutDeepSeek V4 Flash 0731 output price fell 44%, $0.180 to $0.100 per million tokens.
price cutDeepSeek V4 Pro 0423 input price fell 33%, $1.04 to $0.698 per million tokens.
price cutDeepSeek V4 Pro 0423 output price fell 33%, $2.08 to $1.40 per million tokens.
price cutQwen3.6 27B input price fell 50%, $0.600 to $0.300 per million tokens.
price cutQwen3.6 27B output price fell 44%, $3.60 to $2.00 per million tokens.
price cutDeepSeek V4 Flash 0423 input price fell 9%, $0.089 to $0.081 per million tokens.
price cutDeepSeek V4 Flash 0423 output price fell 9%, $0.177 to $0.162 per million tokens.
price riseMoonshotAI Kimi Latest input price rose 2%, $2.50 to $2.55 per million tokens.
price cutMoonshotAI Kimi Latest output price fell 9%, $14.00 to $12.75 per million tokens.
price riseGLM Latest output price rose 13%, $3.50 to $3.96 per million tokens.

Released in the last 90 days

2 of these 20 arrived cheaper than anything matching their score. The rest launched already beaten.

2026-10-07Claude Haiku 5.51303$0.215 / 1k10 hosts
2026-10-06Mistral Large 41249$1.00 / 1k3 hosts
2026-09-29GPT-6.1 Sol1372$4.43 / 1k8 hosts
2026-09-28Claude Sonnet 5.51382$4.03 / 1k8 hosts
2026-09-23Solar Mini 41158$0.088 / 1k2 hostsbest value
2026-09-22Claude Opus 5.51401$8.86 / 1k11 hosts
2026-09-21MiMo-V2.6-Pro1328$0.472 / 1k4 hostsbest value
2026-09-21Grok 4.71272$3.40 / 1k5 hosts
2026-09-02Muse Spark 1.31344$1.98 / 1k1 hosts
2026-09-02Gemini 3.8 Flash1296$1.54 / 1k6 hosts
2026-09-01Claude Fable 5.11344$19.98 / 1k4 hosts
2026-08-26GLM 5.3 Flash1289$0.230 / 1k32 hosts
2026-08-18GLM 5.31295$1.93 / 1k41 hosts
2026-08-13Gemini 3.7 Flash1305$1.54 / 1k6 hosts
2026-08-12Grok 4.61284$3.40 / 1k6 hosts
2026-08-10Solar Pro 41177$0.164 / 1k2 hosts
2026-08-05Muse Spark 1.21310$1.98 / 1k1 hosts
2026-07-31DeepSeek V4 Flash 07311200$0.233 / 1k27 hosts
2026-07-24Claude Opus 51333$11.25 / 1k11 hosts
2026-07-21Gemini 3.6 Flash1287$1.54 / 1k7 hosts

How this is tracked

Every refresh archives a snapshot of each model and each host serving it: price, cached price, context, precision, uptime and host count. The next run diffs against the previous snapshot. A change is reported only where the data moved. Price moves under 2% are treated as noise and ignored.

The precision line is the one worth watching. A host switching a model from fp8 to fp4 keeps the price identical while measurably degrading output, and nobody announces it. See the finding for why that matters, or download the dataset.