Vibe Code Bench 1–100
Building and extending applications across sequences of user requests, beyond a single initial app-generation prompt.
Application developmentVals AI20 models run by Vals AIUpdated 29 September 2026
How it is run
Vals runs the multi-request suite in OpenHands and publishes its overall score with effort settings. Each model’s score retains that evaluator’s harness.
What it cannot tell you
This multi-request suite is distinct from Vibe Code Bench v1.1; scores from the two task sets are not interchangeable.
- Leader
- Claude Opus 5.530.36% · Vals AI's own list
- Top-ten spread
- 12.12%Leader minus tenth place, or last place in smaller fields.
- Median
- 17.88%of the 20 rows published
- No published result
- 26of 46 models on this site
Independent leaderboard · Vals AI
Top 20 of 20 runs by Vals AI, one harness.
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5.5 | 30.36% |
| 2 | Claude Opus 5 | 28.53% |
| 3 | Claude Fable 5.1 | 28.00% |
| 4 | GPT-6 Astra | 27.64% |
| 5 | GPT-5.6 Luna | 22.59% |
| 6 | Muse Spark 1.3 Max | 20.46% |
| 7 | GPT-5.6 Sol | 20.01% |
| 8 | GLM 5.3 | 19.99% |
| 9 | Gemini 3.8 Flash | 18.77% |
| 10 | Kimi K3 | 18.24% |
| 11 | DeepSeek V4 Pro | 17.53% |
| 12 | DeepSeek V4.1 Flash | 16.38% |
| 13 | GLM 5.3 Flash | 16.03% |
| 14 | GPT-5.6 Terra | 14.82% |
| 15 | Grok 4.6 | 14.75% |
| 16 | Claude Sonnet 5 | 13.82% |
| 17 | Qwen3.8 Max | 12.83% |
| 18 | MiniMax M3 | 9.22% |
| 19 | Inkling | 7.33% |
| 20 | Gemini 3.1 Pro | 6.69% |
No published result tracked
26 of the 46 models on this site have no published number on this benchmark, so their cells in the home table stay blank rather than reading as zero.
Step 5 PreviewGPT-6 LunaQwen3.8 2.4T A95BQwen3.8-Flash-NextMiMo-V2.6-FlashQwen3.8 Max (0902)Claude Fable 5Claude Opus 4.8Claude Opus 4.7Gemini 3.7 FlashMuse Spark 1.2GPT-5.4Grok 4.5GPT-5.5DeepSeek V4.1 Flash 1017GLM-LegalClaude Haiku 5.5BeamGemini 4 ArgonClaude Sonnet 5.5GPT-6 SolGPT-6.1 SolGrok 4.7Hunyuan Hy4 PreviewMiMo V2.6 ProMistral Medium 3.5
Read this with
Other versions of the same evaluation. Versions are not comparable: the task set changes between them.
Sources
Every page used for this benchmark, read on 17 October 2026.
- Benchmark site or the page that publishes itwww.vals.ai/benchmarks/vcb-1-100
- Independent resultswww.vals.ai/benchmarks/vcb-1-100