qwen2.5-72b-instruct: benchmark scores, price and ranking
qwen2.5-72b-instruct from Alibaba ranks #711 of 743 LLMs on the LLMs Tiger aggregate benchmark score (42.2), based on 8 benchmark results captured up to 2026-10-13.
- Organisation
- Alibaba
- Type
- Open-weight / local model
- Open weights
- Yes
- Released
- 2024-09-19 (source: epoch)
- Context window
- 32,768 tokens
- Cheapest API price
- $0.120 in / $0.300 out per million tokens (via hyperbolic, observed 2026-10-13T19:47Z)
- Aggregate rank
- #711 of 743 (score 42.2)
Relative to other models, qwen2.5-72b-instruct is strongest on DTBench (#88 of 146) and weakest on WeirdML (#98 of 109).
Across the 8 benchmarks where at least ten models are scored, qwen2.5-72b-instruct is in the top 10% on 0 and in the top half on 0.
Its cheapest listed input price, $0.120 per million tokens, is lower than that of 77% of the 427 models with a listed price.
qwen2.5-72b-instruct benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| DTBench | 62.9% | #88 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| GPQA Diamond | 49.1% | #188 of 275 | Epoch AI Benchmarking Hub | 2026-10-13 |
| OTIS Mock AIME | 8.1% | #141 of 189 | Epoch AI Benchmarking Hub | 2026-10-13 |
| BALROG | 16.2% | #26 of 34 | Epoch AI Benchmarking Hub | 2026-10-13 |
| TheAgentCompany | 5.7% | #10 of 13 | Epoch AI Benchmarking Hub | 2026-10-13 |
| METR time horizon | 5 min | #25 of 31 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 13.4% | #90 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WeirdML | 16.0% | #98 of 109 | Epoch AI Benchmarking Hub | 2026-10-13 |
About qwen2.5-72b-instruct
qwen2.5-72b-instruct is an open-weight model that can run on your own hardware from Alibaba. It was released on 2024-09-19, according to Epoch AI. It accepts up to 32,768 tokens of input at once. The cheapest API listing we track charges $0.120 per million input tokens and $0.300 per million output tokens, through hyperbolic. We hold 8 benchmark results for it across 3 areas, and none is in the top tenth of its board.
Where qwen2.5-72b-instruct stands
On our aggregate score, qwen2.5-72b-instruct is #711 of 743 ranked models, in the lower half. Its score is 42.2; the scale is centred on 50, so it sits 7.8 points below the typical model. That beats 4% of the ranked field. Among the 74 ranked Alibaba models it is #73. The next one up is Qwen2.5-Coder-0.5B-Instruct at 42.6. The next one down is qwen2.5-7b-instruct at 39.9.
qwen2.5-72b-instruct, benchmark by benchmark
qwen2.5-72b-instruct lands below the middle on DTBench: 62.9%, #88 of 146, with 60% of the field scoring higher. Claude Opus 5.5 (max) leads at 98.9%; the closest result above is Gemini 2.0 Flash (Feb 2025) at 63.2%. It covers dynamic tool-use benchmark. The board median is 68.8%, so qwen2.5-72b-instruct is 5.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen2.5-72b-instruct lands below the middle on GPQA Diamond: 49.1%, #188 of 275, with 68% of the field scoring higher. GPT-6 Astra (max) leads at 95.8%; the closest result above is GPT-4o (Aug 2024) at 49.2%. It covers expert-level science questions. The board median is 71.7%, so qwen2.5-72b-instruct is 22.6 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen2.5-72b-instruct lands below the middle on OTIS Mock AIME: 8.1%, #141 of 189, with 74% of the field scoring higher. Claude Fable 5.1 (max) leads at 100.0%; the closest result above is mistral-large-2407 at 8.5%. It covers competition maths (AIME 2024-25 style). The board median is 57.8%, so qwen2.5-72b-instruct is 49.7 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen2.5-72b-instruct lands below the middle on BALROG: 16.2%, #26 of 34, with 74% of the field scoring higher. GPT-6 Astra (max) leads at 68.3%; the closest result above is Llama-3.2-11B-Vision-Instruct at 16.8%. It covers long-horizon game-playing agents. The board median is 27.6%, so qwen2.5-72b-instruct is 11.4 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
qwen2.5-72b-instruct lands below the middle on TheAgentCompany: 5.7%, #10 of 13, with 69% of the field scoring higher. gemini-2.5-flash-preview-09-2025 leads at 41.1%; the closest result above is Llama-3.1-70B-Instruct at 6.9%. It covers work tasks in a simulated software company. The board median is 8.6%, so qwen2.5-72b-instruct is 2.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
METR time horizon is a weak spot. qwen2.5-72b-instruct scores 5 min, #25 of 31, behind 77% of the models measured. The leader, Claude Mythos Preview (Early), reaches 1,045 min. It covers length of task (minutes) done at 50% success. The board median is 23 min, so qwen2.5-72b-instruct is 18 minutes below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
LMCA is a weak spot. qwen2.5-72b-instruct scores 13.4%, #90 of 108, behind 82% of the models measured. The leader, Claude Opus 5.5 (max), reaches 68.2%. It covers language-model capability assessment. The board median is 29.2%, so qwen2.5-72b-instruct is 15.8 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
WeirdML is a weak spot. qwen2.5-72b-instruct scores 16.0%, #98 of 109, behind 89% of the models measured. The leader, GPT-6 Astra (max), reaches 93.3%. It covers unusual machine-learning tasks. The board median is 41.2%, so qwen2.5-72b-instruct is 25.2 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Strengths and weaknesses
Reasoning, maths and science: weak, ahead of 29% of the field on average across 2 benchmarks. Its best board here is GPQA Diamond (#188 of 275) and its weakest is OTIS Mock AIME (#141 of 189). Agents and tool use: weak, ahead of 29% of the field on average across 3 benchmarks. Its best board here is DTBench (#88 of 146) and its weakest is TheAgentCompany (#10 of 13). General knowledge and chat: weak, ahead of 15% of the field on average across 3 benchmarks. Its best board here is METR time horizon (#25 of 31) and its weakest is WeirdML (#98 of 109).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so qwen2.5-72b-instruct costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on Llama 4 Maverick ($0.188) and Trinity Large Thinking ($0.250).
Release history
The Alibaba model released before it was Qwen2.5-72B on 2024-09-19, which scores 50.1. The next one was Qwen2.5-7B on 2024-09-19, at 46.5.
Reading these numbers
The 8 scores on this page come from 1 source: Epoch AI Benchmarking Hub (8). No official leaderboard supplied any of them, so treat them as third-party measurements.
Models ranked nearby
- #708 GPT-4o (Nov 2024) (42.3)
- #709 GPT-4o (Aug 2024) (42.2)
- #710 claude-2.0 (42.2)
- #712 granite-3.1-1b-a400m-base (42.1)
- #713 Llama 3.2 1B (42.0)
- #714 SmolLM2-135M (42.0)
More from Alibaba
- Qwen 3.6 Max (Preview) (#25, 54.1)
- Qwen3.8 Max (xhigh) (#37, 53.2)
- Qwen3.7 Max (#71, 52.3)
- ollama/qwen2.5-coder:32b (#93, 51.9)
- Qwen3 235B A22B diff, no think, Alibaba API (#122, 51.5)
- Qwen2.5-Coder-14B-Instruct (#135, 51.3)
Compare it on the live leaderboard → · How the aggregate score works