GPT-OSS 120B gpt-oss-120b
openai · gpt-oss reasoning flagship efficient
unranked — no quality signals yet
- params
- 117B (A5.1B)
- arch
- moe
- context
- 128k
- license
- apache-2.0 open weights
- released
- Aug 2025
- reasoning
- yes
- train compute
- 4.9e+24 FLOP
- downloads/30d
- 2.5M
# architecture
- layers
- 36
- d_model
- 2880
- heads
- 64q / 8kv
- head dim
- 64
- experts
- top-4 of 128
- vocab
- 201088
- family
- gpt_oss
signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.
# why ranked
Overall: provisional — score 29.2 ● low — a single signal; treat with caution
provisional: too few independent signals for a numbered position — the score is shown, but this model sorts after every ranked model.
| signal | weight | input (0–100) |
|---|---|---|
| aa_intelligence | 0.70 | — |
| bench_composite | 0.30 | 29.2 |
benchmark panel evidence:
| complete panel | score (0–100) |
|---|---|
| aider-polyglot-v1 | 47.9 |
| swe-bench-verified-v1 | 10.4 |
missing signals are dropped and the remaining weights renormalized — never imputed.
Coding: #5 — score 19.5 ● medium — one signal missing, weights renormalized
| signal | weight | input (0–100) |
|---|---|---|
| aa_coding | 0.30 | — |
| aider_polyglot | 0.30 | 45.5 |
| swe_bench_verified | 0.40 | 0.0 |
missing signals are dropped and the remaining weights renormalized — never imputed.
Agentic: #22 — score — ● low — a single signal; treat with caution
| signal | weight | input (0–100) |
|---|---|---|
| swe_bench_verified | 0.60 | 0.0 |
| swe_rebench | 0.40 | — |
missing signals are dropped and the remaining weights renormalized — never imputed.
# trend
methodology changed during this history window; score movement across that boundary is not model movement. See methodology v6.
# benchmarks
# where to run
| provider | quant | ctx | $/M in | $/M out | $/M cache | price src | tps | uptime | ✓ |
|---|---|---|---|---|---|---|---|---|---|
| Featherless | — | — | — | — | via hfrouter | — | — | ||
| PrimeIntellect | fp4 | 128k | $0.02 | $0.07 | $0.02 | via openrouter | — | 98.5% | |
| Near AI | fp4 | 128k | $0.03 | $0.15 | $0.01 | via openrouter | — | 100.0% | |
| CoreWeave | fp4 | 128k | $0.03 | $0.17 | $0.03 | via openrouter | — | 99.9% | |
| Wandb | 131k | $0.03 | $0.17 | — | via litellm | — | — | ||
| DekaLLM | bf16 | 128k | $0.03 | $0.18 | $0.03 | via openrouter | — | 100.0% | |
| AkashML | bf16 | 128k | $0.03 | $0.19 | $0.03 | via openrouter | — | 100.0% | |
| DeepInfra | bf16 | 128k | $0.04 | $0.17 | — | deepinfra | — | 99.9% | ✓ |
| DeepInfra | fp8 | 128k | $0.04 | $0.17 | — | deepinfra | — | 85.5% | ✓ |
| Kyma | 125k | $0.05 | $0.23 | — | via litellm | — | — | ||
| Crusoe | bf16 | 128k | $0.05 | $0.25 | $0.05 | via openrouter | — | 100.0% | |
| Mancer 2 | fp8 | 128k | $0.05 | $0.25 | — | via openrouter | — | 99.9% | |
| Novita | fp4 | 128k | $0.05 | $0.25 | — | novita | — | 99.6% | ✓ |
| Venice | int4 | 125k | $0.07 | $0.30 | — | venice | — | 98.4% | ✓ |
| Amazon Bedrock | unknown | 128k | $0.07 | $0.30 | — | bedrock | — | 100.0% | ✓ |
| Amazon Bedrock | unknown | 128k | $0.07 | $0.30 | — | bedrock | — | 100.0% | ✓ |
| Sail | 128k | $0.06 | $0.40 | — | via litellm | — | — | ||
| Sail Research | fp4 | 128k | $0.06 | $0.40 | $0.03 | via openrouter | — | 100.0% | |
| DigitalOcean | unknown | 125k | $0.06 | $0.42 | $0.01 | via openrouter | — | 100.0% | |
| unknown | 128k | $0.09 | $0.36 | — | via openrouter | — | 62.5% | ||
| Nscale | 128k | $0.10 | $0.40 | — | via hfrouter | 116 | — | ||
| OVHcloud | 128k | $0.09 | $0.47 | — | ovhcloud | 41 | — | ✓ | |
| BaseTen | fp4 | 128k | $0.10 | $0.50 | $0.10 | via openrouter | — | 98.8% | |
| BaseTen | fp4 | 128k | $0.10 | $0.50 | $0.10 | via openrouter | — | 97.8% | |
| Aiand | 128k | $0.15 | $0.60 | — | via litellm | — | — | ||
| Azure | — | $0.15 | $0.60 | — | azure | — | — | ✓ | |
| Bedrock Mantle | 128k | $0.15 | $0.60 | — | via litellm | — | — | ||
| Fireworks | 128k | $0.15 | $0.60 | $0.01 | via requesty | — | — | ||
| Groq | unknown | 128k | $0.15 | $0.60 | — | groq | — | 100.0% | ✓ |
| Nebius | fp4 | 128k | $0.15 | $0.60 | $0.10 | openrouter | — | 100.0% | ✓ |
| Phala | unknown | 128k | $0.15 | $0.60 | — | phala | — | 99.4% | ✓ |
| SiliconFlow | fp8 | 128k | $0.15 | $0.60 | $0.07 | openrouter | — | 85.7% | ✓ |
| Tensormesh | 128k | $0.15 | $0.60 | — | via litellm | — | — | ||
| Together | unknown | 128k | $0.15 | $0.60 | — | via openrouter | — | 100.0% | |
| Zerogpu | 125k | $0.15 | $0.60 | — | via litellm | — | — | ||
| Parasail | fp4 | 128k | $0.10 | $0.75 | $0.06 | via openrouter | — | 100.0% | |
| IBM watsonx | 128k | $0.16 | $0.64 | — | via modelsdev | — | — | ||
| Scaleway | — | $0.17 | $0.68 | — | via hfrouter | 136 | — | ||
| Mara | unknown | 128k | $0.15 | $0.75 | — | via openrouter | — | 99.5% | |
| SambaNova | unknown | 128k | $0.22 | $0.59 | — | sambanova | — | 99.7% | ✓ |
| Replicate | — | $0.18 | $0.72 | — | via litellm | — | — | ||
| Inferenceapis | 128k | $0.20 | $0.80 | — | via litellm | — | — | ||
| Cerebras | fp16 | 128k | $0.35 | $0.75 | $0.35 | openrouter | — | 100.0% | ✓ |
| Cloudflare | 125k | $0.35 | $0.75 | — | via modelsdev | — | — |
sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price
source aliases
- aa
gpt-oss-120b- aider
gpt-oss-120b (high)- arena
gpt-oss-120b- openrouter
openai/gpt-oss-120b