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. | 0.5B/1.5B/3B/7B/14B/32B/72B | — |
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 | 200k |
| 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. | $0.288DeepInfra | $1.20Z.AI |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | $0.32DeepInfra | $4.00Z.AI |
| $92.00DeepInfra · fp8 | $440Z.AI | |
| These models have no shared benchmark scores. | ||
BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. | — | 31% |
LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | 55.5% | — |
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. | 86.6% | — |
ZClawBenchOpenClaw task performance — mean task score on a zero-to-one scale across 116 OpenClaw tasks | — | 0.564 |
IFEvalAgentic tasks — following verifiable instructions in generated responses | 84.1% | — |
MATHReasoning — accuracy on the original competition mathematics test set | 83.1% | — |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | 71.1% | — |
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. | 95.8% | — |
| Overview | ||
| Company | Qwen | Z.ai |
| Release date | Sep 19 2024 | Mar 15 2026 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
Qwen2.5vsClaude Fable 5.2GLM-5-TurbovsClaude Fable 5.2Qwen2.5vsGPT-6.2 AstraGLM-5-TurbovsGPT-6.2 AstraQwen2.5vsGemini 4 NeonGLM-5-TurbovsGemini 4 NeonQwen2.5vsMuse EmberGLM-5-TurbovsMuse EmberQwen2.5vsGrok 4.7GLM-5-TurbovsGrok 4.7Qwen2.5vsDeepSeek-V4.1-Flash-1017GLM-5-TurbovsDeepSeek-V4.1-Flash-1017Frequently asked questions
Qwen2.5 and GLM-5-Turbo don't publish scores on any of the same benchmarks, so there's no direct head-to-head comparison. Qwen2.5 shipped 542 days before GLM-5-Turbo, so benchmark comparisons should account for the intervening progress.
Context windows are 128k (Qwen2.5) vs 200k (GLM-5-Turbo). Qwen2.5 is open weight, while GLM-5-Turbo is closed.
Direct benchmark comparisons are unavailable — Qwen2.5 and GLM-5-Turbo don't publish scores on any of the same benchmarks.