Get started

gpt-oss-120b (batch)

Sold by 23 sellers, timed at 223 tok/s, scored 3.8 on the Intelligence Index, on 6 Design Arena boards, and with results on 2 openly licensed boards.

retrieved sha256 · 0bb53a77713f

Models/OpenAI/gpt-oss-120b (batch)USD per 1M tokens · as OpenRouter routes it
openai/gpt-oss-120b:batchOpenAI · batch variant · GPT tokenizer
OpenRouter list price · the routed rate, as the catalog carried itretrieved 17 Nov 16:29 UTC from openrouter.ai ↗
Input$0.0296per 1M tokens
Output$0.136per 1M tokens
Cached inputnot listedper 1M tokens
Context131k131,072 tokens
What the list says · the fields OpenRouter publishes for this id12 fields
AuthorOpenAI
Listed
Context131,072 tokens
Max output117,964 tokens
Modalitiestext → text
TokenizerGPT
Reasoningalways on · effort high, medium, low, medium by default
Moderationnot moderated by OpenRouter
Variantbatch
Knowledge cutoff2024-06-30
Other listingbase listing
What the weights say · read from the model's own repository on Hugging Faceretrieved from huggingface.co ↗
Licenceapache-2.0
Parameters117B · 116,829,156,672 declared
Model classGptOssForCausalLM
Downloadnot gated
Published

What sellers charge

23 sellers · 11.67x from cheapest to dearest

One tick per seller, on a log scale of input price per million from $0.03 at AkashML to $0.35 at PrimeIntellect. The $0.0296 listed is the cheapest on offer; 23 sellers are dearer. A faint tick declares no number format. Every seller is named on the endpoint ledger below.

The same sellers grouped by the number format each declares, on the same scale. The fp4 group alone spans 99% of the range, so the cheap end is not cheap for its number format. A declaration is checked by nobody: this groups the rows and certifies none.

Measured speed

225,515 requests · 23 endpoints · 30-minute window
Throughput223 tok/sMedian output rate, weighted across the 23 endpoints serving it by the requests each answered.the speed board
First token339msMedian wait before the first token, over the same window and weighted the same way.how Vitna reads a figure
Endpoint spread19 to 635The slowest and the fastest route serving this model, in tokens per second. Speed is a property of the route, not of the weights.the board, fastest first

OpenRouter's own timing of its routed traffic, weighted by requests, over one 30-minute window read . Every route is on the endpoint ledger below.

Every endpoint

23 priced · 23 timed · cheapest first
  1. AkashMLakashml/fp4$0.03$0.187$0.03fp4118k106 tok/s77420ms32,678 req100%
  2. CoreWeavecoreweave/fp4$0.03$0.17$0.03fp4118k155 tok/s120197ms2,779 req96.87%
  3. DekaLLMdekallm/bf16$0.03$0.18$0.03bf16118k113 tok/s52.6611ms239 req70.60%
  4. Near AInear-ai/fp4$0.03$0.15$0.015fp4118k56 tok/s38586ms35,628 req100%
  5. Venicevenice/int4$0.03$0.15$0.03int433k37 tok/s19891ms2,073 req98.21%
  6. DeepInfradeepinfra/bf16$0.037$0.17$0.0185bf16118k19 tok/s14568ms8,837 req73.05%
  7. Crusoecrusoe/bf16$0.05$0.25$0.05bf16118k223 tok/s136219ms39,374 req100%
  8. DigitalOceandigitalocean$0.06$0.42$0.012not declared4k64 tok/s39892ms125 req100%
  9. Mancer 2mancer/fp8$0.085$0.375$0.085fp8118k101 tok/s45828ms30 req99.70%
  10. Googlegoogle-vertex/global$0.09$0.36not listednot declared118k226 tok/s161262ms30,018 req99.93%
  11. BaseTenbaseten/fp4/us$0.10$0.50$0.10fp4115k133 tok/s109239ms92 req99.77%
  12. BaseTenbaseten/fp4$0.10$0.50$0.10fp4115k137 tok/s96161ms2,929 req99.34%
  13. Parasailparasail/fp4$0.10$0.75$0.055fp4118knot matchednonenonenot matched100%
  14. SambaNovasambanova$0.14$0.95not listednot declared118k206 tok/s76.41.34s164 req99.40%
  15. Amazon Bedrockamazon-bedrock$0.15$0.60not listednot declared118k291 tok/s175517ms2,935 req99.92%
  16. Amazon Bedrockamazon-bedrock/eu-west-1$0.15$0.60not listednot declared118k164 tok/s118434ms374 req100%
  17. Groqgroq$0.15$0.60$0.075not declared66k357 tok/s241201ms39,406 req99.02%
  18. Nebiusnebius/fp4$0.15$0.60$0.10fp4118k181 tok/s92349ms1,158 req99.88%
  19. Parasailparasail/fast$0.15$0.75not listedfp4118knot matchednonenonenot matched100%
  20. SiliconFlowsiliconflow/fp8$0.15$0.60$0.075fp88k32 tok/s231.11s278 req100%
  21. Togethertogether$0.15$0.60not listednot declared118k99 tok/s57209ms6,781 req100%
  22. Cerebrascerebras/fp16$0.35$0.75$0.35fp1641k635 tok/s241258ms19,008 req100%
  23. PrimeIntellectprimeintellect/fp4$0.35$0.75$0.315fp466k178 tok/s90.3419ms142 req98.81%
  24. Parasailnot matchednot matchednot matchednot matchedno offer116 tok/s88364ms48 reqno offer
  25. Parasailnot matchednot matchednot matchednot matchedno offer147 tok/s118368ms419 reqno offer

Each track runs from a route's p90, the rate nine in ten of its requests met or beat, up to its median. Weighted by requests, the figure above reports where the traffic went rather than what one route does: 18 of the 23 timed routes are slower than it, and the middle route runs at 137 tok/s. The widest median-to-p90 gap is SambaNova's, 2.7x.

Parasail publishes 2 offers and had 2 routes timed. The two documents share no route id, so those offers and routes are printed apart rather than paired.

Uptime is OpenRouter's own figure for each seller over the trailing day, read once a day, and the column prints the latest reading, taken . 25 of the 30 sellers with a reading fell under 99% on at least one of the 23 days held, 23 Oct 2026 to 21 Nov 2026: Google at 1.45% on 4 Nov 2026, DeepInfra (deepinfra/bf16) at 31.41% on 17 Nov 2026, DeepInfra (deepinfra/fp8) at 34.35% on 10 Nov 2026, SambaNova at 69.68% on 26 Oct 2026, and 21 more. Source: OpenRouter's public endpoints API, one reading a day for each seller: 1,211 sellers carried one on the latest of the 23 days held.

Source: OpenRouter's public endpoints API. 1257 sellers across 359 models when the mirror ran, 186 of them served by more than one. List prices, not what anybody paid; a discount is the seller's own. Read . Speed is OpenRouter's own timing over one 30-minute window, read .

What each seller accepts

21 parameters · 23 sellers
  1. include_reasoningevery seller
  2. max_tokensevery seller
  3. reasoningevery seller
  4. reasoning_effortevery seller
  5. temperatureevery seller
  6. top_pevery seller
  7. stop20 of 23not DigitalOcean, Google, and SiliconFlow
  8. tool_choice20 of 23not DigitalOcean, Google, and SiliconFlow
  9. tools20 of 23not DigitalOcean, Google, and SiliconFlow
  10. response_format18 of 23
  11. structured_outputs18 of 23
  12. frequency_penalty17 of 23
  13. top_k17 of 23
  14. presence_penalty16 of 23
  15. seed14 of 23
  16. repetition_penalty12 of 23
  17. logit_bias10 of 23
  18. logprobs9 of 23
  19. top_logprobs9 of 23
  20. min_p5 of 23
  21. top_a2 of 23only Mancer 2 and PrimeIntellect

6 of these 21 are accepted by every one of the 23 sellers that publish a list; the other 15 are declared by some sellers and not others, so what a request sending one of them gets depends on which seller serves it. Each list is the seller's own declaration in OpenRouter's endpoints API, read , and nothing here tested one.

Price position

342 priced models

Blended $0.056 per million tokens, 3 in to 1 out: cheaper than 332 of the 342 priced models in the catalog. Each bar is one stretch of the catalog on a logarithmic price scale; the mark is this listing.

Measured quality

145 models with an index · 132 on a design arena board

Each mark is one board this model stands on, placed at the share of that board it scores above, and led to its number in the list under the rule. The columns and the ledger below carry the figures.

Intelligence Index3.8Position 57 of 58, above 2% of the models scored.the intelligence leaderboard
Coding Index30.4Position 84 of 118, above 29% of the models scored.the coding leaderboard
Agentic Index3.7Position 67 of 94, above 29% of the models scored.the agentic leaderboard
  1. 3D scenes · Models881 Elowin rate 29.4% · source rank 131the board, 113 ranked
  2. Code · Models951 Elowin rate 33.4% · source rank 139the board, 120 ranked
  3. Data visualisation · Models984 Elowin rate 43.6% · source rank 122the board, 110 ranked
  4. Game development · Models981 Elowin rate 40.5% · source rank 129the board, 120 ranked
  5. UI components · Models915 Elowin rate 35.7% · source rank 133the board, 116 ranked
  6. Websites · Models959 Elowin rate 32.5% · source rank 142the board, 126 ranked
Epoch Capabilities Index · a fit by a third party over every benchmark this model has results onthe whole index, with intervals
Index140.2138.3 to 142.0
Position127 of 258above 51% of the models placed
Results behind it13benchmark results
As Epoch names itgpt-oss-120breleased 5 Aug 2025

Position 127 of the 258 models Epoch has placed, which is not the population of either board above, on a scale anchored at Claude 3.5 Sonnet 130 and GPT-5 150. Epoch states that absolute values are meaningless by themselves and that comparisons between models are what the index is for, that values are not linearly related to benchmark accuracy, and that a highly specialised model may score low while being capable in its own domain.

Benchmark results

2 of 6 boards · 3 results
Epoch AI Benchmarking Hub · publisher-run, CC-BY-4.01 of 22 tasks · 149 runs scored on the task
TaskEffortScore
Mystery Game Puzzleshigh0
medium0.02

Mean accuracy, 0 to 1, on the task the row names, as Epoch AI runs it and averaged over repeated runs. All eighteen of their tasks, including the held-out FrontierMath tiers, named as they name them. The model id carries the effort setting the run used. Epoch AI Benchmarking Hub names these runs gpt-oss-120b_high and gpt-oss-120b_medium, and the effort beside each score says which run it is. Read from epoch.ai ↗, which dates its own data . Epoch AI Benchmarking Hub data, by Epoch AI, under CC BY 4.0.

Aider polyglot · submitted, Apache-2.01 of 1 task · 69 runs scored on the task · at high effort
TaskScore
polyglot pass rate41.80

Percent of 225 Exercism exercises solved with a second attempt allowed, higher is better. The id carries the edit format, because the publisher says a model alone does not identify a run. Aider polyglot names this run gpt-oss-120b (high) (diff). Read from raw.githubusercontent.com ↗, which dates its own data . Aider polyglot benchmark, Apache-2.0, by the Aider project. Results are submitted by contributors rather than produced by the publisher.

2 of the 6 openly licensed boards Vitna mirrors carry a result for gpt-oss-120b (batch), each printed as its publisher published it and none added to another. MTEB, LiveBench, HELM and Vectara hallucination leaderboard have not published one under a name that matches its listing, and names are matched to the catalog exactly and never guessed. A task that is a link opens its column on /capability, with every run scored on it. Every board, with its licence and what its publisher measures, is on /benchmarks.

Sources

Source: OpenRouter (openrouter.ai/models), as of 17 Nov 16:29 UTC. sha256 0bb53a77713f over 853,270 bytes.
Catalog: OpenRouter's public model list.
Intelligence, coding and agentic indexes: Artificial Analysis, as published by OpenRouter.
Elo, win rate and rank: Design Arena, as published by OpenRouter.
Source: each model's own Hugging Face repository (huggingface.co), as of 20 Nov 2026. 162 repositories read, 4 did not answer.
Licence, parameter count, model class and gating: each model's own Hugging Face repository, the one OpenRouter names for the listing.
A licence is the tag the uploader put on the repository, read from nothing else and verified by nobody.
A parameter count is the total the repository's safetensors files declare; for a mixture of experts that is every expert, not the share one token runs through.
Source: the Epoch AI Capabilities Index (epoch.ai/eci), as of 20 Nov 2026. 258 models placed over 2917 benchmark results.
Epoch AI Benchmarking Hub data, by Epoch AI, under CC BY 4.0.
Vitna ran none of these evaluations and adjusted none of these figures.