Vals Public Benefits Bench v1.1 (Public Benefits Bench v1.1)
We show this table for reference; we do not rank on it.
Can AI help people navigate SNAP benefits?
Evaluation results published by Vals AI. Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.
Public Benefits Bench v1.1 score on Public Benefits Bench v1.1 — October 12, 2026
We mirror the published public benefits bench v1.1 score view for Public Benefits Bench v1.1. Claude Opus 5 leads the public snapshot at 76.93%, followed by Claude Fable 5.1 (74.90%) and Muse Wonder Full (73.75%). We do not use these results to rank models overall.
Claude Opus 5
Anthropic
Claude Fable 5.1
Anthropic
Muse Wonder Full
Meta
max reasoning
50 modelsAgenticCurrentDisplay onlyUpdated October 12, 2026
Public Benefits Bench v1.1 score table (50 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 5Anthropic | 76.93% | Not reported | Closed |
| 2 | Claude Fable 5.1Anthropic | 74.90% | Not reported | Closed |
| 3 | Muse Wonder FullMetamax reasoning | 73.75% | Not reported | Closed |
| 4 | Claude Opus 5.5Anthropic | 70.64% | Not reported | Closed |
| 5 | Claude Fable 5Anthropic | 70.43% | Not reported | Closed |
| 6 | Muse WonderMetamax reasoning | 70.23% | Not reported | Closed |
| 7 | Gemini 4 ArgonGooglehigh reasoning | 69.76% | Not reported | Closed |
| 8 | MiMo-V2.6-ProXiaomi | 68.94% | Not reported | Open |
| 9 | Hy4 previewTencent | 68.61% | Not reported | Open |
| 10 | GLM-5.3Z.AImax reasoning | 68.54% | Not reported | Open |
| 11 | Muse Spark 1.2Metaxhigh reasoning | 68.47% | Not reported | Closed |
| 12 | Kimi K3Moonshot AImax reasoning | 68.20% | Not reported | Open |
| 13 | Claude Opus 4.8Anthropic | 68.13% | Not reported | Closed |
| 14 | MiMo-V2.6-FlashXiaomi | 67.59% | Not reported | Open |
| 15 | Claude Sonnet 5.5Anthropic | 67.19% | Not reported | Closed |
| 16 | Qwen3.8 MaxAlibaba | 67.12% | Not reported | Open |
| 17 | Grok 4.6xAIhigh reasoning | 66.85% | Not reported | Closed |
| 18 | GPT-5.6 SolOpenAImax reasoning | 66.51% | Not reported | Closed |
| 19 | Ember-1Fireworks | 66.17% | Not reported | Closed |
| 20 | Claude Sonnet 5Anthropic | 66.03% | Not reported | Closed |
| 21 | Grok 4.7xAIxhigh reasoning | 65.63% | Not reported | Closed |
| 22 | Gemini 3.8 FlashGooglehigh reasoning | 65.29% | Not reported | Closed |
| 23 | DeepSeek V4.1 Flash 1017DeepSeekmax reasoning | 64.28% | Not reported | Open |
| 24 | MiniMax M3MiniMax | 64.14% | Not reported | Open |
| 25 | Claude Haiku 5.5Anthropic | 63.19% | Not reported | Closed |
| 26 | DeepSeek V4 Pro 0813DeepSeekmax reasoning | 62.92% | Not reported | Open |
| 27 | Claude Sonnet 4.6Anthropic | 62.45% | Not reported | Closed |
| 28 | GPT-5.6 TerraOpenAImax reasoning | 62.38% | Not reported | Closed |
| 29 | GLM-5.1Z.AI | 61.84% | Not reported | Open |
| 30 | Muse Spark 1.1Metaxhigh reasoning | 61.30% | Not reported | Closed |
| 31 | GPT-5.6 LunaOpenAImax reasoning | 61.16% | Not reported | Closed |
| 32 | GPT-5.5OpenAIxhigh reasoning | 60.89% | Not reported | Closed |
| 33 | Gemini 3.5 FlashGooglehigh reasoning | 59.47% | Not reported | Closed |
| 34 | Inkling-SmallThinking Machines Lab0.99 reasoning | 59.41% | Not reported | Open |
| 35 | GPT-6.1 SolOpenAImax reasoning | 59.34% | Not reported | Closed |
| 36 | InklingThinking Machines Lab0.99 reasoning | 58.46% | Not reported | Open |
| 37 | Step 5 PreviewStepFun | 58.12% | Not reported | Open |
| 38 | GPT-6 LunaOpenAImax reasoning | 57.65% | Not reported | Closed |
| 39 | Kimi K2.6Moonshot AI | 56.63% | Not reported | Open |
| 40 | GPT-6 SolOpenAImax reasoning | 56.63% | Not reported | Closed |
| 41 | Gemini 3.6 FlashGooglehigh reasoning | 56.56% | Not reported | Closed |
| 42 | Claude Haiku 4.5Anthropic | 54.26% | Not reported | Closed |
| 43 | Gemini 3.1 Pro PreviewGooglehigh reasoning | 53.79% | Not reported | Unknown |
| 44 | Gemini 3.5 Flash-LiteGooglehigh reasoning | 51.83% | Not reported | Closed |
| 45 | Grok 4.3xAIhigh reasoning | 51.69% | Not reported | Closed |
| 46 | Ling 3.0 Flash 2607Ant | 50.74% | Not reported | Unknown |
| 47 | Mercury 2.5Inceptionhigh reasoning | 45.26% | Not reported | Closed |
| 48 | Grok 4.1 Fast (Reasoning)xAI | 44.79% | Not reported | Closed |
| 49 | Laguna M.1Poolside | 43.84% | Not reported | Closed |
| 50 | Laguna XS.2Poolside | 41.41% | Not reported | Open |
Public Benefits Bench v1
BenchLM mirrors this Vals AI leaderboard for reference only. It is excluded from BenchLM overall and category rankings.
| Source rank | Source model or configuration | Public Benefits Bench v1 score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Fable 5Anthropic | 71.65% | Vals AIOriginal result | Not reported | Closed |
| 2 | Claude Opus 4.8Anthropic | 62.11% | Vals AIOriginal result | Not reported | Closed |
| 3 | MiniMax M3MiniMax | 60.69% | Vals AIOriginal result | Not reported | Open |
| 4 | Claude Sonnet 4.6Anthropic | 58.52% | Vals AIOriginal result | Not reported | Closed |
| 5 | Gemini 3.5 FlashGoogle | 57.98% | high reasoningOriginal result | Not reported | Closed |
| 6 | GLM-5.1Z.AI | 57.92% | Vals AIOriginal result | Not reported | Open |
| 7 | DeepSeek V4 Pro 0813DeepSeek | 57.58% | max reasoningOriginal result | Not reported | Open |
| 8 | GPT-5.5OpenAI | 57.24% | xhigh reasoningOriginal result | Not reported | Closed |
| 9 | Kimi K2.6Moonshot AI | 53.38% | Vals AIOriginal result | Not reported | Open |
| 10 | Gemini 3.1 Pro PreviewGoogle | 50.88% | high reasoningOriginal result | Not reported | Unknown |
| 11 | Grok 4.3xAI | 50.07% | high reasoningOriginal result | Not reported | Closed |
| 12 | Claude Haiku 4.5Anthropic | 49.53% | Vals AIOriginal result | Not reported | Closed |
| 13 | Grok 4.1 Fast (Reasoning)xAI | 44.32% | Vals AIOriginal result | Not reported | Closed |
How Public Benefits Bench v1.1 is shown here
BenchLM mirrors the public Vals AI Public Benefits Bench v1.1 leaderboard captured from https://www.vals.ai/benchmarks/public-benefits-bench and updated by Vals on October 12, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.
Public Benefits Bench v1.1 is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.
Snapshot
The published Public Benefits Bench v1.1 snapshot places Claude Opus 5 first at 76.93%. The third row is 3.18 points behind. The broader top-10 range is 8.39 points, so many of the published results sit in a relatively narrow band.
50 models have been evaluated on Public Benefits Bench v1.1. The benchmark falls in the Agentic category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. Public Benefits Bench v1.1 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About Public Benefits Bench v1.1
Year
2026
Tasks
SNAP public-benefits navigation tasks
Format
Accuracy score
Difficulty
Public-benefits policy navigation
BenchLM mirrors the public Vals AI Public Benefits Bench v1.1 leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.
Freshness and provenance
Version
Public Benefits Bench v1.1 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does Public Benefits Bench v1.1 measure?
Can AI help people navigate SNAP benefits?
Which model leads the published Public Benefits Bench v1.1 snapshot?
Claude Opus 5 currently leads the published Public Benefits Bench v1.1 snapshot with 76.93% public benefits bench v1.1 score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on Public Benefits Bench v1.1?
The October 12, 2026 snapshot contains 50 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.