Coding / Benchmark profile
Vals.ai benchmark for evaluating whether models can build complete web applications from natural language specifications in a production-like development environment.
Observed results
20 ranked models · higher is better · labels show rank and score
| Model | Rank | Model ID | Provider | Score | Unit |
|---|---|---|---|---|---|
| Claude Opus 4.7 | #1 | claude-opus-4-7 | Anthropic | 71.003 | percent |
| GPT-5.5 | #2 | gpt-5-5 | OpenAI | 69.847 | percent |
| GPT-5.4 | #3 | gpt-5-4 | OpenAI | 67.421 | percent |
| GPT-5.3-Codex | #4 | gpt-5-3-codex | OpenAI | 61.767 | percent |
| Claude Opus 4.6 | #5 | claude-opus-4-6 | Anthropic | 57.573 | percent |
| GPT-5.2 | #6 | gpt-5-2 | OpenAI | 53.499 | percent |
| Claude Sonnet 4.6 | #7 | claude-sonnet-4-6 | Anthropic | 51.476 | percent |
| DeepSeek V4 Pro | #8 | deepseek-v4-pro | DeepSeek | 49.931 | percent |
| Gemini 3.5 Flash | #9 | gemini-3-5-flash | 48.683 | percent | |
| GPT-5.4 mini | #10 | gpt-5-4-mini | OpenAI | 47.969 | percent |
| GPT-5.2 Codex | #11 | gpt-5-2-codex | OpenAI | 37.912 | percent |
| Kimi K2.6 | #12 | kimi-k2-6 | Moonshot AI | 37.891 | percent |
| GLM-5.1 | #13 | glm-5-1 | Z.AI | 31.456 | percent |
| MiniMax M2.7 | #14 | minimax-m2-7 | MiniMax | 27.037 | percent |
| GPT-5.4 nano | #15 | gpt-5-4-nano | OpenAI | 26.097 | percent |
| Qwen3.6 Plus | #16 | qwen3-6-plus | Alibaba Cloud | 25.564 | percent |
| GPT-5.1 | #17 | gpt-5-1 | OpenAI | 24.606 | percent |
| GLM-5 (Reasoning) | #18 | glm-5-reasoning | Z.AI | 23.359 | percent |
| Claude Sonnet 4.5 Thinking | #19 | claude-sonnet-4-5-thinking | Anthropic | 22.621 | percent |
| GPT-5.1-Codex-Max | #20 | gpt-5-1-codex-max | OpenAI | 22.168 | percent |
| Claude Opus 4.5 Thinking | #21 | claude-opus-4-5-thinking | Anthropic | 20.63 | percent |
| Gemini 3 Flash | #22 | gemini-3-flash | 20.204 | percent | |
| GPT-5 | #23 | gpt-5 | OpenAI | 20.088 | percent |
| Muse Spark | #24 | muse-spark | Meta | 19.674 | percent |
| Kimi K2.5 | #25 | kimi-k2-5 | Moonshot AI | 17.536 | percent |
| Qwen3.5 Plus | #26 | qwen3-5-plus | Alibaba Cloud | 15.738 | percent |
| MiniMax M2.5 | #27 | minimax-m2-5 | MiniMax | 14.852 | percent |
| GPT-5 mini | #28 | gpt-5-mini | OpenAI | 14.171 | percent |
| GPT-5.1-Codex | #29 | gpt-5-1-codex | OpenAI | 13.115 | percent |
| Claude Haiku 4.5 Thinking | #30 | claude-haiku-4-5-thinking | Anthropic | 11.393 | percent |
| DeepSeek V3.2 (Thinking) | #31 | deepseek-v3-2-thinking | DeepSeek | 5.108 | percent |
| Grok 4.20 | #32 | grok-4-20 | xAI | 4.064 | percent |
| Qwen3 Max | #33 | qwen3-max | Alibaba Cloud | 3.506 | percent |
| GLM-4.6 | #34 | glm-4-6 | Z.AI | 3.09 | percent |
| Grok 4.1 Fast (Reasoning) | #35 | grok-4-1-fast-reasoning | xAI | 1.2 | percent |
| Gemini 2.5 Pro | #36 | gemini-2-5-pro | 0.4 | percent | |
| Gemini 3.1 Flash-Lite | #37 | gemini-3-1-flash-lite | 0 | percent | |
| Grok 4 Fast (Reasoning) | #38 | grok-4-fast-reasoning | xAI | 0 | percent |
One best compatible score per canonical product · higher is better
| Rank | Model | Weights | Evidence | Score |
|---|---|---|---|---|
| 01 | Claude Opus 4.7Anthropic | Closed weights | Source carrier | 71.0% |
| 02 | GPT-5.5OpenAI | Closed weights | Source carrier | 69.8% |
| 03 | GPT-5.4OpenAI | Closed weights | Source carrier | 67.4% |
| 04 | GPT-5.3-CodexOpenAI | Closed weights | Source carrier | 61.8% |
| 05 | Claude Opus 4.6Anthropic | Closed weights | Source carrier | 57.6% |
| 06 | GPT-5.2OpenAI | Closed weights | Source carrier | 53.5% |
| 07 | Claude Sonnet 4.6Anthropic | Closed weights | Source carrier | 51.5% |
| 08 | DeepSeek V4 ProDeepSeek | Open weights | Source carrier | 49.9% |
| 09 | Gemini 3.5 FlashGoogle | Closed weights | Source carrier | 48.7% |
| 10 | GPT-5.4 miniOpenAI | Closed weights | Source carrier | 48.0% |
Each distinct published measurement is retained. Same-snapshot canonical and configuration projections appear once.
| Published model / configuration | Score | Evidence & protocol | Source & dates |
|---|---|---|---|
Exact BenchLM registry variant Claude Opus 4.7; bulk export does not retain a complete upstream harness configuration.Claude Opus 4.7Exact identitySource label without a registered configuration ID Canonical product: claude-opus-4-7 | 71.0% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant GPT-5.5; bulk export does not retain a complete upstream harness configuration.GPT-5.5Exact identitySource label without a registered configuration ID Canonical product: gpt-5-5 | 69.8% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant GPT-5.4; bulk export does not retain a complete upstream harness configuration.GPT-5.4Exact identitySource label without a registered configuration ID Canonical product: gpt-5-4 | 67.4% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant GPT-5.3 Codex; bulk export does not retain a complete upstream harness configuration.GPT-5.3-CodexExact identitySource label without a registered configuration ID Canonical product: gpt-5-3-codex | 61.8% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant Claude Opus 4.6; bulk export does not retain a complete upstream harness configuration.Claude Opus 4.6Exact identitySource label without a registered configuration ID Canonical product: claude-opus-4-6 | 57.6% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant GPT-5.2; bulk export does not retain a complete upstream harness configuration.GPT-5.2Exact identitySource label without a registered configuration ID Canonical product: gpt-5-2 | 53.5% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant Claude Opus 4.6 (Adaptive); bulk export does not retain a complete upstream harness configuration.Claude Opus 4.6Exact identityclaude-opus-4-6-max Canonical product: claude-opus-4-6 | 53.5% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant Claude Sonnet 4.6; bulk export does not retain a complete upstream harness configuration.Claude Sonnet 4.6Exact identitySource label without a registered configuration ID Canonical product: claude-sonnet-4-6 | 51.5% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant DeepSeek V4 Pro (Max); bulk export does not retain a complete upstream harness configuration.DeepSeek V4 ProExact identitydeepseek-v4-pro-max Canonical product: deepseek-v4-pro | 49.9% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
Exact BenchLM registry variant Gemini 3.5 Flash; bulk export does not retain a complete upstream harness configuration.Gemini 3.5 FlashExact identitySource label without a registered configuration ID Canonical product: gemini-3-5-flash | 48.7% | Source carriersource-checkedVersion & system2026 Source-native system | BenchLM public datasets — 2026-08-01Observed Checked |
From result to context
Vals.ai benchmark for evaluating whether models can build complete web applications from natural language specifications in a production-like development environment.
The summary shows one best compatible source result per canonical product. Versions, effort settings and execution systems remain attached to the underlying records.
Scores from different versions or harnesses may not be interchangeable. The published source rows preserve those distinctions and their original units. A source result is not automatically an input to a current index.
This catalogue entry preserves available source evidence. Current indices admit only their specifically reviewed tracks and configurations.
Contamination risk: Unknown. Lifecycle: Active.
