gpt-oss-120b
gpt-oss-120b is served by 24 companies at prices ranging from $0.03 to $0.35 per million input tokens, a 11.7× difference for identical weights.
- cheapest host · Venice
- $0.03
- price spread across hosts
- 12×
- companies serving it
- 24
- intelligence index
- 11.6
Is it worth it?
Beaten on both axes. DeepSeek V4 Flash 0423 scores 24.4 against 11.6 and costs 1.7× less for the same work.
Every host, compared
The same weights, priced and configured differently by 24 companies. Precision matters: 4-bit builds measurably degrade output and cost the same.
| Host | $/1M in | $/1M out | Cached | Context | Precision | Uptime |
|---|---|---|---|---|---|---|
| Venicecheapest | $0.030 | $0.150 | $0.030 | 128K | int4 | 99.7% |
| CoreWeave | $0.030 | $0.170 | $0.030 | 131K | fp4 | 100% |
| DekaLLM | $0.030 | $0.180 | $0.030 | 131K | bf16 | 100% |
| AkashML | $0.030 | $0.187 | $0.030 | 131K | bf16 | 100% |
| DeepInfra | $0.037 | $0.170 | n/a | 131K | bf16 | 99.1% |
| Crusoe | $0.050 | $0.250 | $0.050 | 131K | bf16 | 99.2% |
| Mancer 2 | $0.050 | $0.250 | n/a | 131K | fp8 | 99.9% |
| Novita | $0.050 | $0.250 | n/a | 131K | fp4 | 99.7% |
| DigitalOcean | $0.060 | $0.420 | $0.012 | 128K | unknown | 100% |
| $0.090 | $0.360 | n/a | 131K | unknown | 79.1% | |
| BaseTen | $0.100 | $0.500 | $0.100 | 128K | fp4 | 100% |
| BaseTen | $0.100 | $0.500 | $0.100 | 128K | fp4 | 99.7% |
| Parasail | $0.100 | $0.750 | $0.055 | 131K | fp4 | 100% |
| SambaNova | $0.140 | $0.950 | n/a | 131K | unknown | 99.1% |
| Amazon Bedrock | $0.150 | $0.600 | n/a | 131K | unknown | 100% |
| Nebius | $0.150 | $0.600 | n/a | 131K | fp4 | 100% |
| Amazon Bedrock | $0.150 | $0.600 | n/a | 131K | unknown | 100% |
| DeepInfra | $0.150 | $0.600 | n/a | 131K | bf16 | 99.9% |
| SiliconFlow | $0.150 | $0.600 | $0.075 | 131K | fp8 | 72.7% |
| Phala | $0.150 | $0.600 | n/a | 131K | unknown | 100% |
| Together | $0.150 | $0.600 | n/a | 131K | unknown | 93.6% |
| Groq | $0.150 | $0.600 | $0.075 | 131K | unknown | 100% |
| Mara | $0.150 | $0.750 | n/a | 131K | unknown | 100% |
| Cerebras | $0.350 | $0.750 | $0.350 | 131K | fp16 | 100% |
7 of 24 hosts serve this model at 4-bit precision (Venice, CoreWeave, Novita, BaseTen, BaseTen, Parasail, Nebius), at the same price as full-precision hosts. See why this matters.
What it costs to run
- Marketing emails
- $0.233
- per 1,000 · short brief in, one email out
- Support replies
- $0.285
- per 1,000 · ticket plus help-centre context
- Lead enrichment
- $0.282
- per 1,000 · company page in, fields out
- Long-form drafts
- $1.07
- per 1,000 · outline in, full draft out
- Questions on docs
- $2.22
- per 1,000 · one large document, many questions
- Agent / tool loop
- $1.94
- 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
- 131K · 197 pages
- Context range across hosts
- 128K to 131K
- Max output
- 118K
- Cache discount
- not published
- Input types
- text
- Tool use
- supported
- Reasoning
- always on, billed as output
- Arena Elo
- 960 across 6 categories
- Agentic Elo
- n/a
- AA Intelligence
- 11.6
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.