Kimi K2.6
Kimi K2.6 is served by 17 companies at prices ranging from $0.46 to $1.09 per million input tokens, a 2.3× difference for identical weights.
- cheapest host · Inceptron
- $0.46
- price spread across hosts
- 2.3×
- companies serving it
- 17
- intelligence index
- 27
Is it worth it?
Beaten on both axes. GLM 5.3 Flash scores 41.8 against 27 and costs 6.6× less for the same work.
Every host, compared
The same weights, priced and configured differently by 17 companies. Precision matters: 4-bit builds measurably degrade output and cost the same.
| Host | $/1M in | $/1M out | Cached | Context | Precision | Uptime |
|---|---|---|---|---|---|---|
| Inceptroncheapest | $0.465 | $2.45 | $0.087 | 262K | int4 | 99.9% |
| Chutes | $0.500 | $2.85 | $0.050 | 262K | int4 | 100% |
| DigitalOcean | $0.570 | $2.40 | $0.114 | 262K | unknown | 100% |
| StreamLake | $0.598 | $2.52 | $0.101 | 256K | fp8 | 97.9% |
| CoreWeave | $0.650 | $3.41 | $0.150 | 262K | fp4 | 99.9% |
| Crusoe | $0.700 | $3.50 | $0.350 | 262K | bf16 | 99.7% |
| DeepInfra | $0.750 | $3.50 | $0.150 | 262K | fp4 | 91.7% |
| Venice | $0.750 | $3.50 | $0.160 | 256K | int4 | 53.2% |
| Parasail | $0.750 | $3.50 | $0.160 | 262K | int4 | 100% |
| SiliconFlow | $0.770 | $3.40 | $0.140 | 262K | fp8 | 100% |
| Novita | $0.800 | $3.40 | $0.160 | 262K | unknown | 99.3% |
| GMICloud | $0.855 | $3.60 | $0.144 | 262K | fp8 | n/a |
| Baidu | $0.950 | $4.00 | $0.160 | 262K | fp4 | 100% |
| AtlasCloud | $0.950 | $4.00 | $0.160 | 262K | int4 | 100% |
| Cloudflare | $0.950 | $4.00 | $0.160 | 262K | unknown | 100% |
| Moonshot AI | $0.950 | $4.00 | $0.160 | 262K | int4 | 99.9% |
| Phala | $1.09 | $4.60 | $0.370 | 262K | unknown | 99.3% |
9 of 17 hosts serve this model at 4-bit precision (Inceptron, Chutes, CoreWeave, DeepInfra, Venice, Parasail, Baidu, AtlasCloud, Moonshot AI), at the same price as full-precision hosts. See why this matters.
What it costs to run
- Marketing emails
- $1.53
- per 1,000 · short brief in, one email out
- Support replies
- $1.52
- per 1,000 · ticket plus help-centre context
- Lead enrichment
- $2.02
- per 1,000 · company page in, fields out
- Long-form drafts
- $6.90
- per 1,000 · outline in, full draft out
- Questions on docs
- $6.80
- per 1,000 · one large document, many questions
- Agent / tool loop
- $6.40
- per 1,000 · long context reused each step
Priced on the median host. Change the workload on the leaderboard to see how the ranking moves.
Specification
- Context, most hosts
- 262K · 393 pages
- Context range across hosts
- 256K to 262K
- Max output
- 236K
- Cache discount
- 80% off cached input
- Input types
- text, image
- Tool use
- supported
- Reasoning
- optional
- Arena Elo
- 1243 across 8 categories
- Agentic Elo
- 1187
- AA Intelligence
- 27
Other models with the same host spread
Host pricing and precision read live from OpenRouter’s endpoint API, compiled 12 October 2026. Free to quote with attribution. See all findings.