Skip to content

Benchmark

Finance Agent v2

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models measured
30
Scores recorded
45
Sources
2
IQ weight
1.55

Score distribution

46.9%
Mean
48.4%
Median
8.9%
Std dev
32.1%
10th percentile
57.9%
90th percentile
27.9%
Min
27.9%61.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.55 for chat models, measured over 30 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Gemini 3.8 Flashfinancial analyst tasks · pass@1· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.61.4%
2Ling 3.0 Flash Fin· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim59.8%
3Gemini 3.5 FlashGemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.57.9%
4Gemini 3.5 FlashOpus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.57.9%
5Gemini 3.5 Flash· LLM Stats data57.9%
6Muse Spark 1.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim57.2%
7Muse Spark 1.1Meta Model API · xhigh · Vals AI result· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.57.2%
8Claude Fable 5· LLM Stats data56.3%
9Mistral Large 4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim54.7%
10Kimi K3max effort · temp 1.0 · top_p 1.0· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.54.4%
11Claude Opus 4.8· LLM Stats data53.9%
12Claude Opus 4.8max · Vals AI result· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.53.9%
13Claude Opus 4.8Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.53.9%
14GPT 5.5Gemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.8%
15GPT 5.5Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.8%
16GPT 5.5xhigh · Vals AI result· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.8%
17GPT 5.5· LLM Stats data51.8%
18Claude Opus 4.7· LLM Stats data51.5%
19Claude Opus 4.7Gemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.5%
20Claude Opus 4.7Opus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51.5%
21Claude Sonnet 4.6· LLM Stats data51%
22Claude Sonnet 4.6Gemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.51%
23Qwen3.7 Max· LLM Stats data48.4%
24MiniMax M3· LLM Stats data48.3%
25GPT-5.4 mini· LLM Stats data45.4%
26Kimi K2.6· LLM Stats data44.9%
27GLM 5.1· LLM Stats data44.8%
28Gemini 3.1 Pro PreviewGemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.43%
29Gemini 3.1 Pro Previewhigh · Vals AI result· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.43%
30Gemini 3.1 Pro PreviewOpus 4.8 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.43%
31Gemini 3.1 Pro Preview· LLM Stats data43%
32Gemini 3 FlashGemini 3.5 Flash Model Card · May 2026· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.42.6%
33Gemini 3 Flash· LLM Stats data42.6%
34MiMo V2.5 Pro· LLM Stats data41.5%
35Qwen3.6 Plus· LLM Stats data40.9%
36GPT 5.4 Nano· LLM Stats data38.2%
37Qwen3.7 Plus· LLM Stats data38.2%
38Grok 4.3· LLM Stats data37.7%
39Nemotron 3 Ultra· LLM Stats data37.5%
40MiMo V2.5· LLM Stats data36.7%
41Mistral Medium 3.5· LLM Stats data32.1%
42Claude Haiku 4.5· LLM Stats data31%
43Gemini 3.1 Flash Lite· LLM Stats data30%
44Grok 4.20reasoning, 0309, beta· LLM Stats data28.5%
45MiniMax M2.7· LLM Stats data27.9%

Data credits: Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.