Compare AI models
| Specifications | ||
ParametersThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. | 21B | — |
Context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider. | 128k | 128k |
| API pricingUSD per 1M tokens · lower wins | ||
Cheapest inputLowest input rate across every provider on OpenRouter, the lab included. The cheapest endpoint may run a quantised build or a shorter context — see "Available from" on the model page. | $0.018AkashML | — |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | $0.09AkashML | — |
| $8.10Darkbloom · fp8 | — | |
| Benchmarks | ||
MMLUGeneral knowledge — A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. | 85.3% | 81.9% |
SWE-Bench VerifiedCoding — Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. | 60.7% | — |
HumanEvalFunction synthesis — 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. | — | 75.7% |
IFEvalAgentic tasks — following verifiable instructions in generated responses | — | 75.2% |
Humanity's Last Exam · no toolsMultidisciplinary reasoning — Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. | 10.9% | — |
Humanity's Last Exam · with toolsMultidisciplinary reasoning — Humanity's Last Exam — extremely hard expert questions across many subjects. “With tools” means the AI is allowed to search the web or run code while answering. Higher is better. | 17.3% | — |
MATHReasoning — accuracy on the original competition mathematics test set | — | 57.9% |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | 92.1% | — |
AIME 2025Reasoning — single-answer accuracy on the 2025 invitational mathematics exam without tools | 91.7% | — |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 71.5% | — |
MMMLUKnowledge — average accuracy on multilingual translations of MMLU | 75.7% | — |
GSM8KGrade-school math — Grade-school maths word problems that take a few steps of arithmetic to work through. It separated the models of 2022 and 2023 sharply, then saturated. One caveat on the historical numbers: OpenAI included part of the GSM8K training set in GPT-4's pre-training mix, so GPT-4's score is not a clean few-shot result. Higher is better. | — | 90.9% |
HealthBench · full setHealthcare — physician rubric score on the full set of the original HealthBench evaluation | 42.5% | — |
HealthBench · hardHealthcare — physician rubric score on the hard subset of the original HealthBench evaluation | 10.8% | — |
HealthBench · consensusHealthcare — physician rubric score on the consensus subset of the original HealthBench evaluation | 82.6% | — |
| Overview | ||
| Company | OpenAI | Z.ai |
| Release date | Aug 5 2025 | Jun 5 2024 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
gpt-oss-20bvsClaude Fable 5.2GLM-4-AirvsClaude Fable 5.2gpt-oss-20bvsGemini 4 HydrogenGLM-4-AirvsGemini 4 Hydrogengpt-oss-20bvsMuse EmberGLM-4-AirvsMuse Embergpt-oss-20bvsGrok 4.7GLM-4-AirvsGrok 4.7gpt-oss-20bvsDeepSeek-V4.1-Flash-1017GLM-4-AirvsDeepSeek-V4.1-Flash-1017gpt-oss-20bvsMistral Large 4GLM-4-AirvsMistral Large 4Frequently asked questions
gpt-oss-20b leads GLM-4-Air on 1 of the 1 benchmark they both report (MMLU). GLM-4-Air shipped 426 days before gpt-oss-20b, so benchmark comparisons should account for the intervening progress.
Context windows are 128k (gpt-oss-20b) vs 128k (GLM-4-Air). gpt-oss-20b is open weight, while GLM-4-Air is closed.
On MMLU, gpt-oss-20b leads at 85.3% vs GLM-4-Air at 81.9%.