Providers / openai/gpt-oss-120b
openai/gpt-oss-120b as each provider declares it
6 of the 24 endpoints declare at least one material setting that differs from the official reference. Declared values, not measured behaviour.
Reference. Context 131,072 tokens, from openai/gpt-oss-120b config.json at b5c939de8f (max_position_embeddings). Quantization: MXFP4 on the MoE experts, bf16 elsewhere.
Provider rows: OpenRouter providers page, the endpoint JSON. Read 2026-10-19.
Endpoints
| Provider | Quantization | Context | Max output | Tools / reasoning | Differs from the reference |
|---|---|---|---|---|---|
| AkashML akashml/bf16 | bf16 (higher precision) | 131,072 | 117,964 | yes / yes | none declared |
| Amazon Bedrock amazon-bedrock | none (not declared) | 131,072 | 117,964 | no / yes | declares no tools parameter; the official model supports tool calling |
| Amazon Bedrock amazon-bedrock/eu-west-1 | none (not declared) | 131,072 | 117,964 | no / yes | declares no tools parameter; the official model supports tool calling |
| BaseTen baseten/fp4 | fp4 (official) | 128,072 | 115,264 | yes / yes | declares a 128,072-token context; the reference config says 131,072 (98%) |
| BaseTen baseten/fp4/us | fp4 (official) | 128,072 | 115,264 | yes / yes | declares a 128,072-token context; the reference config says 131,072 (98%) |
| Cerebras cerebras/fp16 | fp16 (higher precision) | 131,072 | 40,960 | yes / yes | none declared |
| CoreWeave coreweave/fp4 | fp4 (official) | 131,072 | 117,964 | yes / yes | none declared |
| Crusoe crusoe/bf16 | bf16 (higher precision) | 131,072 | 117,964 | yes / yes | none declared |
| DeepInfra deepinfra/bf16 | bf16 (higher precision) | 131,072 | 117,964 | yes / yes | none declared |
| DekaLLM dekallm/bf16 | bf16 (higher precision) | 131,072 | 117,964 | yes / yes | none declared |
| DigitalOcean digitalocean | none (not declared) | 128,000 | 4,096 | no / yes | declares a 128,000-token context; the reference config says 131,072 (98%)declares no tools parameter; the official model supports tool callingdeclares at most 4,096 output tokens, which the reasoning tokens also count against |
| Google google-vertex/global, status -2 | none (not declared) | 131,072 | 117,964 | no / yes | declares no tools parameter; the official model supports tool calling |
| Groq groq | none (not declared) | 131,072 | 65,536 | yes / yes | none declared |
| Mancer 2 mancer/fp8 | fp8 (higher precision) | 131,072 | 117,964 | yes / yes | none declared |
| Mara mara, status -2 | none (not declared) | 131,072 | 117,964 | yes / yes | none declared |
| Near AI near-ai/fp4 | fp4 (official) | 131,072 | 117,964 | yes / yes | none declared |
| Nebius nebius/fp4 | fp4 (official) | 131,072 | 117,964 | yes / yes | none declared |
| Novita novita/fp4, status -2 | fp4 (official) | 131,072 | 32,768 | yes / yes | none declared |
| Parasail parasail/fp4 | fp4 (official) | 131,072 | 117,964 | yes / yes | none declared |
| Phala phala, status -2 | none (not declared) | 131,072 | 32,768 | yes / yes | none declared |
| SambaNova sambanova | none (not declared) | 131,072 | 117,964 | yes / yes | none declared |
| SiliconFlow siliconflow/fp8 | fp8 (higher precision) | 131,072 | 8,192 | no / yes | declares no tools parameter; the official model supports tool calling |
| Together together | none (not declared) | 131,072 | 117,964 | yes / yes | none declared |
| Venice venice/int4 | int4 (differs) | 128,000 | 32,768 | yes / yes | declares a 128,000-token context; the reference config says 131,072 (98%)declares int4 weights; the official release is MXFP4 on the MoE experts, bf16 elsewhere |
Changes since tracking began
None yet. A row is recorded only when a declared value or the reference moves.
How this is read
Once a day the tracker reads OpenRouter's public endpoint list for each model (one row per provider deployment: declared context length, declared maximum output, declared quantization, and the request parameters it accepts) and the official repository's config.json at its head commit. Both reads are anonymous. A new row is recorded only when either moved, so the history below is a list of real changes.
Every value in the tables is what a provider declares to OpenRouter. No request is sent to any model, so nothing here says how an endpoint behaves: one that declares a 131,072-token context may still cut off earlier, and one that declares the official quantization may still serve something else.
A quantization the provider does not declare cannot be compared, and is counted as undisclosed rather than as matching. A quantization of higher precision than the official weights is shown and not counted as a difference.
Material means one of: a declared context under 95% of the reference config, a declared quantization other than the official release, no tools parameter on a model that supports tool calling, no reasoning control on a model that exposes one, or a reasoning model capped under 8,192 output tokens. A context between 95% and 100% of the reference is listed and not counted as material.
Sampling defaults, chat templates and stop tokens are not in the endpoint metadata and are not compared here. Moonshot's K2 Vendor Verifier (github.com/MoonshotAI/K2-Vendor-Verifier) measures tool-call behaviour across providers for Kimi K2 by sending requests; this page does not.
OpenRouter lists only the providers that serve through it. Nothing here ranks providers, and a provider with no listed difference has not been shown to match the reference, only to declare nothing that differs.