SWE-bench Pro benchmark
33 AI models ranked on SWE-bench Pro. Each row shows the model’s score next to its cheapest API price per 1M tokens where it has one, so you can weigh quality against cost.
Top model today
Qwen3.8 2.4T A95B (Alibaba) leads at 68%, from $2.00 per 1M input.
Data via Scale AI SWE-bench Pro →Models ranked
33
on this benchmark
Top score
68%
current leader
Type
Single eval
benchmark
Updated
29 Aug 2026
last source fetch
Cost vs SWE-bench Pro
31 priced modelsEach labelled name links to that model’s page. Curated to the cost-efficiency frontier, top scorers and cheapest so labels stay readable; the full field is in the leaderboard table below (and via “select all”). Reasoning models (extended “thinking” before answering) carry a ring around the dot.
Models ranked on SWE-bench Pro
highest score first · 33 models| # | ||||||
|---|---|---|---|---|---|---|
| 1 | Qwen3.8 2.4T A95B |  Alibaba | 68% | $2.00 | $6.00 | |
| 2 | Hy4 preview | Other | 66% | $0.83 | $2.50 | |
| 3 | Qwen3.8-Flash-Next | Alibaba | 63% | n/a | n/a | |
| 4 | GLM 5.2 |  Zhipu | 62% | $0.61 | $1.98 | |
| 5 | Qwen3.8 27B | Alibaba | 62% | $0.40 | $2.55 | |
| 6 | Laguna S 2.1 | Other | 59% | $0.09 | $0.18 | |
| 7 | MiniMax M3 |  MiniMax | 59% | $0.23 | $0.96 | |
| 8 | Kimi K2.6 |  Moonshot | 59% | $0.65 | $3.40 | |
| 9 | GLM 5p1 | Zhipu | 58% | $1.40 | $4.40 | |
| 10 | Hy3 | Other | 58% | $0.13 | $0.53 | |
| 11 | MiMo-V2.5-Pro |  Xiaomi | 57% | $0.43 | $0.87 | |
| 12 | Ling-3.0-flash | Other | 57% | $0.02 | $0.06 | |
| 13 | Step 3.7 Flash | Other | 56% | $0.20 | $1.15 | |
| 14 | MiniMax M2.7 | MiniMax | 56% | $0.25 | $0.55 | |
| 15 | MiMo-V2.5 | Xiaomi | 56% | $0.14 | $0.28 | |
| 16 | Inkling Small | Other | 56% | $0.45 | $1.20 | |
| 17 | DeepSeek V4 Pro 0423 |  DeepSeek | 55% | $0.43 | $0.87 | |
| 18 | MiniMax M2.5 | MiniMax | 55% | $0.27 | $1.08 | |
| 19 | Inkling | Other | 54% | $0.95 | $4.05 | |
| 20 | Qwen3.6 27B | Alibaba | 54% | $0.15 | $0.50 | |
| 21 | Muse Glimmer 30B | Other | 51% | $0.30 | $1.20 | |
| 22 | Kimi K2.5 | Moonshot | 51% | $0.45 | $2.25 | |
| 23 | Qwen3.6 35B A3B | Alibaba | 50% | $0.10 | $0.45 | |
| 24 | Laguna M.1 | Other | 49% | $0.20 | $0.40 | |
| 25 | Laguna XS 2.1 | Other | 48% | $0.06 | $0.12 | |
| 26 | Qwen3 Coder Next | Alibaba | 44% | $0.12 | $0.80 | |
| 27 | North Mini Code |  Cohere | 40% | n/a | n/a | |
| 28 | Qwen3 Coder 480B A35b Instruct | Alibaba | 39% | $0.38 | $1.50 | |
| 29 | MiniMax M2.1 | MiniMax | 37% | $0.30 | $1.20 | |
| 30 | Kimi K2 Instruct | Moonshot | 28% | $0.50 | $2.00 | |
| 31 | Qwen3 235B A22B | Alibaba | 21% | $0.18 | $0.54 | |
| 32 | gpt-oss-120b |  OpenAI | 16% | $0.03 | $0.17 | |
| 33 | GLM 4.6 | Zhipu | 10% | $0.43 | $1.75 |
Frequently asked questions
What is the SWE-bench Pro benchmark?
SWE-bench Pro (Scale AI) is a harder, contamination-resistant successor to SWE-bench, built from larger and more recent real-world software tasks across many repositories. The model must generate a patch that passes the project's tests. Scored as the percentage of tasks resolved, with frontier models still well short of saturation.
Which AI model scores highest on SWE-bench Pro?
Qwen3.8 2.4T A95B (Alibaba) leads with 68%, from $2.00 per 1M input tokens.
How many models are ranked on SWE-bench Pro?
33 models carry a SWE-bench Pro score. Scores come from Hugging Face leaderboards; where a model has a live API price it is verified against LiteLLM and OpenRouter, and models with no current price are listed without one.
Other benchmarks
compare the same models on a different evalEvery weekday
AI moves fast. Here's your debrief.
News, analysis, tools, and more.
For people who build with AI