Benchmarks / OpenHands Index

Reported by OpenHands Index

OpenHands Index

Building a library from its specification until its test suite passes.

Data older than 90 days: last updated 30 Jun 2026

Results dated
30 Jun 2026
Models
34
Unit
% resolved
Licence
Apache License 2.0

OpenHands Index in brief

What it measures. OpenHands Index is a third-party benchmark, shown here with credit. Measured here: Building a library from its specification until its test suite passes.

Latest results. , 34 models ranked.

Top 3 on Greenfield (Commit0) (% resolved, higher is better):

  1. Claude Fable 5 (Anthropic): 62.5%
  2. Claude Opus 4.8 (Anthropic): 62.5%
  3. Claude Opus 4.6 (Anthropic): 56.2%

The first 2 are too close to tell apart, so read the order among them as a tie.

How to cite. OpenHands Index, via Spring Prompt, results dated 30 June 2026. https://springprompt.com/benchmarks/openhands-index These results belong to OpenHands Index, not to us. Licence: Apache License 2.0. Credit OpenHands Index first.

Greenfield (Commit0):

Top 15 of 34 results · % resolved, higher is better · ≈ cannot be told apart from the leader. Choose a model to highlight it.Clear highlight

  1. 1≈ Claude Fable 5Anthropic 62.5%
  2. 1≈ Claude Opus 4.8Anthropic 62.5%
  3. 3 Claude Opus 4.6Anthropic 56.2%
  4. 3 Claude Opus 4.7Anthropic 56.2%
  5. 3 GPT-5.4OpenAI 56.2%
  6. 6 Claude Sonnet 4.6Anthropic 50.0%
  7. 6 GPT-5.2OpenAI 50.0%
  8. 8 GPT-5.2-CodexOpenAI 43.8%
  9. 8 GPT-5.5OpenAI 43.8%
  10. 10 Claude Opus 4.5Anthropic 37.5%
  11. 10 Gemini 3.5 FlashGoogle 37.5%
  12. 10 GLM-5.1Z.ai 37.5%
  13. 13 GLM-5Z.ai 31.2%
  14. 14 Qwen3.6-PlusAlibaba 25.0%
  15. 14 Qwen3-Coder-NextAlibaba 25.0%

Full results

OpenHands Index: greenfield (Commit0), % resolved, higher is better
#ModelGreenfield (Commit0)
% resolved, higher is better
Price
$ per million tokens, in / out
1≈ Claude Fable 5Anthropic
62.5%
$10 / $50
1≈ Claude Opus 4.8Anthropic
62.5%
$5 / $25
3 Claude Opus 4.6Anthropic
56.2%
$5 / $25
3 Claude Opus 4.7Anthropic
56.2%
$5 / $25
3 GPT-5.4OpenAI
56.2%
$2.50 / $15
6 Claude Sonnet 4.6Anthropic
50.0%
$3 / $15
6 GPT-5.2OpenAI
50.0%
$1.75 / $14
8 GPT-5.2-CodexOpenAI
43.8%
$1.75 / $14
8 GPT-5.5OpenAI
43.8%
$5 / $30
10 Claude Opus 4.5Anthropic
37.5%
$5 / $25
10 Gemini 3.5 FlashGoogle
37.5%
$1.50 / $9
10 GLM-5.1Z.ai
37.5%
$1.33 / $4.18
13 GLM-5Z.ai
31.2%
$1 / $3.20
14 Qwen3.6-PlusAlibaba
25.0%
$0.33 / $1.95
14 Qwen3-Coder-NextAlibaba
25.0%
$0.12 / $0.80
14 DeepSeek-V3.2 (reasoning on)DeepSeek
25.0%
$0.28 / $0.42
14 Gemini 3.1 Pro PreviewGoogle
25.0%
$2 / $12
14 Gemini 3 Pro PreviewGoogle
25.0%
–
14 MiniMax-M3MiniMax
25.0%
$0.30 / $1.20
14 Kimi K2.6Moonshot AI
25.0%
$0.95 / $4
21 Gemini 3 Flash PreviewGoogle
18.8%
$0.50 / $3
21 MiniMax-M2.1MiniMax
18.8%
–
21 MiniMax-M2.7MiniMax
18.8%
$0.30 / $1.20
21 Kimi K2.5Moonshot AI
18.8%
$0.57 / $2.85
25 Qwen3.5-FlashAlibaba
12.5%
$0.065 / $0.26
25 Claude Sonnet 4.5Anthropic
12.5%
$3 / $15
25 Trinity Large ThinkingArcee AI
12.5%
–
25 DeepSeek-V4-Pro (0423)DeepSeek
12.5%
$1.42 / $2.83
25 MiniMax-M2.5MiniMax
12.5%
$0.29 / $1.20
25 Kimi K2 ThinkingMoonshot AI
12.5%
$0.60 / $2.50
25 Nemotron 3 Super 120B A12BNVIDIA
12.5%
$0.08 / $0.45
25 GLM-4.7Z.ai
12.5%
$0.60 / $2.20
33 Nemotron 3 Nano 30B A3BNVIDIA
6.2%
$0.06 / $0.24
34 Qwen3-Coder-480B-A35BAlibaba
0.0%
$0.30 / $1

Swipe the table sideways for more columns.

Source: OpenHands Index, via Spring Prompt, springprompt.com/benchmarks/openhands-index. Test your own prompt on these models →

Ranks follow the score as shown, so equal numbers share a rank. Results as published by OpenHands Index; we do not re-run them.

Progress over time

Greenfield (Commit0), plotted at each model's release date, higher is better. Each model counts once, at its best setting. Scores are today's, not launch day's; 1 model without a known release date is left off. Hover or tap a mark for its model.

Best score so far

Every model at its release date. The line moves each time a new release beats every earlier one, and each new best is named where there is room.

-20%0%20%40%60%80% Oct 20252026AprJulOct Qwen3-Coder-480B-A35BClaude Sonnet 4.5Gemini 3 Pro PreviewClaude Opus 4.5GPT-5.2Claude Opus 4.6Claude Opus 4.8

Best score by lab

Each lab's best so far, stepping whenever one of its releases beats its own earlier models. The 6 labs with the best scores today; 4 others not drawn.

AnthropicOpenAIGoogleZ.aiAlibabaMoonshot AI

-20%0%20%40%60%80% Oct 20252026AprJulOct AnthropicOpenAIGoogleZ.aiAlibabaMoonshot AI

What it measures

Building a library from its specification until its test suite passes.

What it does not measure

Not the model alone: the score is for the model inside the OpenHands agent, at the SDK version shown.

OpenHands Index by the OpenHands contributors, Apache License 2.0.