Co-Design Bench
Agents revise existing applications from multimodal user requests. A judge checks browser and repository evidence against 2,331 expert criteria across 273 tasks.
The evaluation uses a judge-then-refute procedure. Scores depend on the judge and execution environment.
- Made by
- Xilin Wei and collaborators
- Released
- October 2026
- Kind
- Web and app front ends
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- Terminal
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents, Vision, Audio and speech
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
In the authors’ words
Evaluating human-AI co-design requires testing whether coding agents can translate real users' multimodal requests into working application revisions. We introduce Co-Design Bench, a reference-free benchmark of 273 tasks curated from 37,463 deduplicated queries across 26,314 real user coding sessions. Its two tracks, Multimodal Editing (ME) and Repository Editing (RE), cover localized edits with optional image, audio, video, or document inputs and coordinated repository revisions. Tasks span three capability dimensions: Grounded Understanding, UI Transformation, and Functional Implementation. The tasks pair normalized requests and runnable pre-edit applications with 2,331 hidden expert-authored acceptance criteria. Our judge-then-refute protocol grades criteria using source and browser evidence, challenges positive verdicts with a separate LLM verifier, and audits visual and interaction claims without gold code or target screenshots. Under a shared agentic harness, the strongest of ten model endpoints passes an average of 69.9% of criteria per task but fully satisfies only 26.4% of tasks. All ten endpoints achieve lower criterion completion on repository edits spanning three or four change types than on those spanning one or two. Co-Design Bench provides a testbed for studying the gap between partial requirement satisfaction and complete task success under agentic execution for real-user application revisions.
GPT-6 Astra leads with 69.9% among models released by September 2026
Best score per model from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- Claude Opus 5.5
- and 13 more
- 1 GPT-6 Astra 69.9%
- 2 Claude Opus 5 66.1%
- 3
Qwen3.8 Max (0902) 65.3%
- 4
Qwen 3.8 27B 64.4%
- 5 Kimi K3 62.9%
- 6 GPT-5.6 Sol 61.1%
- 7 Claude Opus 4.8 58.1%
- 8 GLM-5.3-Flash 57.3%
- 9 GLM-5.2 56.7%
- 10 Grok 4.6 51.1%
Source: The Co-Design Bench paper
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 69.9% | Claude Code | Sep 2026 | Official leaderboard |
| 2 | Claude Opus 5 | 66.1% | Claude Code | Jul 2026 | Official leaderboard |
| 3 | | 65.3% | Claude Code | Sep 2026 | Official leaderboard |
| 4 | | 64.4% | Claude Code | Aug 2026 | Official leaderboard |
| 5 | Kimi K3 | 62.9% | Claude Code | Jul 2026 | Official leaderboard |
| 6 | GPT-5.6 Sol | 61.1% | Claude Code | Jul 2026 | Official leaderboard |
| 7 | Claude Opus 4.8 | 58.1% | Claude Code | May 2026 | Official leaderboard |
| 8 | GLM-5.3-Flash | 57.3% | Claude Code | Aug 2026 | Official leaderboard |
| 9 | GLM-5.2 | 56.7% | Claude Code | Jun 2026 | Official leaderboard |
| 10 | Grok 4.6 | 51.1% | Claude Code | Aug 2026 | Official leaderboard |
Rankings by part
We read 2 parts from the benchmark’s paper: Graded completion and Full pass. Each ranks models on one part of the test. GPT-6 Astra leads both. Arrows show how far a model moved from its overall rank among the models we track.
Graded completion
10 models we track have a score here.
- 1 GPT-6 Astra 75.3%
- 2 Claude Opus 5 71.9%
- 3
Qwen3.8 Max (0902) 71.6%
- 4
Qwen 3.8 27B 70.9%
- 5 Kimi K3 69.5%
- 6 GPT-5.6 Sol 67.9%
- 7 Claude Opus 4.8 64.9%
- 8 GLM-5.3-Flash 63.9%
- 9 GLM-5.2 63.7%
- 10 Grok 4.6 58%
Full pass
10 models we track have a score here.
- 1 GPT-6 Astra 26.4%
- 2 Claude Opus 5 18.3%
- 3
Qwen3.8 Max (0902) 16.5%
- 4 Kimi K3 ↑1 15.8%
- 5
Qwen 3.8 27B ↓1 15%
- 6 GPT-5.6 Sol 14.3%
- 7 GLM-5.3-Flash ↑1 10.6%
- 8 Claude Opus 4.8 ↓1 10.3%
- 9 GLM-5.2 9.50%
- 10 Grok 4.6 6.60%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The Co-Design Bench paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Co-Design Bench leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Coding
- SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
- LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests.
DeepSeek-V4-Pro 93.5%
- SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
- DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
- SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases.
Seed2.1 Pro 59.8%
- Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance.
DeepSeek V4.1 Flash 3,471
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (10 entries).
- Pass Models scored 10 models we track.
- Pass Newest score The newest is from September 2026, for GPT-6 Astra.
- Note Frontier models Only 1 of the top 10 has a score. The strongest tested is GPT-6 Astra (#4).
- Pass Headroom The best is 69.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on codesignbench.github.io: 10 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.