Vals AI evaluates sustained multi-turn software development.
Each task starts with a complete, passing web application from Vibe Code Bench v1.1, then asks a model to implement a sequence of dependent product requests while preserving existing behavior. This is distinct from Vibe Code Bench v1.1 zero-to-one application generation. The verified source edition is 1.0.
Compare recorded scores and measured cost.
Current recorded best scores, ordered by model release date.
Hover or tap to inspect. Keyboard: ← / → to move, Home / End to jump, Esc to reset.
Muse Wonder Full: 34.0%, released Oct 12, 2026
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 34.0% | $19.54 | 1.37B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 30.4% | $44.12 | 2.1B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 28.5% | $41.49 | 2.86B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 28.0% | $149.66 | 3.85B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 27.6% | $84.72 | 1.93B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 26.6% | $13.72 | 1B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 22.6% | $4.91 | 3.74B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Thinking High | 21.2% | $18.66 | 750M |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 20.5% | $5.60 | 956M |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 20.4% | $8.30 | 2.61B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 20.0% | $47.43 | 3.2B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 20.0% | $8.24 | 1.72B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Thinking High | 18.8% | $18.91 | 3.49B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 18.2% | $9.40 | 1.15B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 17.5% | $5.89 | 1.63B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 16.4% | $0.69 | 1.62B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 16.0% | $0.46 | 927M |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 14.8% | $33.33 | 2.46B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Thinking High | 14.8% | $26.27 | 1.05B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Thinking High | 14.3% | $6.75 | — |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Max | 13.8% | $71.15 | 3.61B |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Effort not stated | 12.8% | $9.68 | 950M |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Effort not stated | 9.2% | $3.03 | 787M |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Effort not stated | 7.3% | $5.00 | 914M |
| Evaluation setting | Score | Cost / task | Total tokens |
|---|---|---|---|
| Thinking High | 6.7% | $3.36 | 276M |
All results are from the official leaderboard · All results use tools.
The dataset license covers the test itself; each score source sets its own terms.
Published by Vals AI · No license recorded; check the publisher’s terms
Unless a license is listed, each source’s own terms apply.
Published by Vals AI
Source: Vals AI. © 2026 Vals AI. All rights reserved. No affiliation or endorsement.