AlibabaQwen2.5 72B
Qwen2.5 72B is an Alibaba open-weights model, released 19 Sept 2024. Its best configuration ranks #602 of 785 on the BenchLeader Index at 42.2 ±3.7, in the lower half. The ± is the point: 286 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in instruction following (46) and lowest in coding (37). At $0.485 per million tokens blended it is cheaper than most ranked models. Output speed of 27 tokens per second puts it in the slowest quarter, with a first answer in 4.0 s. Last measured 22 Oct 2026.
- Blended price
- $0.485/M
- $0.480 in · $0.500 out
- Output speed
- 27 tok/s
- measured by Artificial Analysis
- First answer
- 3.99 s
- first token 0.61 s
- Context
- 33k
- Full answer
- 22 s
- median, reasoning included
- Cost per run
- –
- Released
- 19 Sept 2024
- Overall index42
- Reasoning41
- Coding37
- Agents & tools38
- Maths44
- Knowledge38
- Instruction following46
- Human preference46
- Composite38
How its index has moved
8 Sept 2026 to 21 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| GPQA Diamond | 49.1%#217 | Epoch AI Benchmarking Hub | |
| Arena Hard Prompts | 1317#246 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 49.1%#433 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 3.6%#589 | Artificial Analysis | |
| GPQA Diamond (HELM)not in index | 42.6%#44 | HELM Capabilities | |
| CritPt | 0.0%#315 | Artificial Analysis | |
| LMCAnot in index | 13.4%#174 | Epoch AI Benchmarking Hub | |
| DTBenchnot in index | 62.9%#172 | Epoch AI Benchmarking Hub | |
| ForecastBenchnot in index | 61.6%#63 | ForecastBench | |
| HellaSwagnot in index | 84.8%#11 | Epoch AI Benchmarking Hub | |
| BIG-Bench Hardnot in index | 79.8%#6 | Epoch AI Benchmarking Hub | |
| WinoGrandenot in index | 82.3%#11 | Epoch AI Benchmarking Hub | |
| ARC (AI2 Challenge)not in index | 94.5%#3 | Epoch AI Benchmarking Hub | |
| PIQAnot in index | 82.6%#18 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| WeirdML | 16.0%#145 | WeirdML | |
| Arena Coding | 1355#239 | Arena (LMArena) |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench Hard | 4.5%#284 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 34.5%#238 | Artificial Analysis | |
| BALROGnot in index | 16.2%#29 | BALROG | |
| TheAgentCompanynot in index | 5.7%#11 | TheAgentCompany | |
| METR Time Horizonsnot in index | 35.8%#33 | METR |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| OTIS Mock AIME | 8.1%#232 | Epoch AI Benchmarking Hub | |
| MATH Level 5 | 63.2%#45 | Epoch AI Benchmarking Hub | |
| Omni-MATH (HELM) | 33.0%#38 | HELM Capabilities | |
| Arena Maths | 1296#231 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -53.1#416 | Artificial Analysis | |
| MMLU-Pro (HELM)not in index | 63.1%#44 | HELM Capabilities | |
| MMLUnot in index | 85.3%#8 | Epoch AI Benchmarking Hub | |
| TriviaQAnot in index | 71.9%#27 | Epoch AI Benchmarking Hub | |
| AA-Omniscience: accuracynot in index | 17.5%#364 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 14.5%#347 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 36.9%#295 | Artificial Analysis | |
| IFEval (HELM) | 80.6%#46 | HELM Capabilities | |
| WildBench (HELM) | 80.2%#31 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1292#246 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1303#254 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1254#263 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1299#242 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1317#244 | Arena (LMArena) |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | 129.0#150 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 7.7#460 | Artificial Analysis | |
| HELM Capabilities meannot in index | 59.9%#44 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 56.4%#7 | Scale AI SEAL |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| DeepInfra | 36 tok/s | 0.39 s | $0.360 | $0.400 | 33k | fp8 |
Price history
The price above, per 1M tokens, every day it changed.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0003 | – | 14.9 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0061 | – | 25.8 s |
| Code edit | 6,000 / 1,500 | $0.0036 | – | 58.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.031 | – | 2.5 min |
| Structured extraction | 2,000 / 200 | $0.0011 | – | 11.3 s |
More from Alibaba
See also
Questions people ask
- What are Qwen2.5 72B's benchmark scores?
- As of 25 Oct 2026, Qwen2.5 72B scores 42.2 ±3.7 on the BenchLeader Index, strongest in instruction following (46) and weakest in coding (37). That rests on 16 independent benchmark results, and every score on this page links to its publisher.
- How does Qwen2.5 72B rank?
- Qwen2.5 72B is #602 of 785 ranked configurations on 25 Oct 2026. 286 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Qwen2.5 72B cost per 1M tokens?
- As of 25 Oct 2026, Qwen2.5 72B costs $0.480 per 1M input tokens and $0.500 per 1M output tokens, $0.485 blended (three input tokens per output token). The list price last changed on 6 Oct 2026.
- When was Qwen2.5 72B released?
- Alibaba released Qwen2.5 72B on 19 Sept 2024.
Data as of 25 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Qwen2.5 72B: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/qwen2-5-72b.