Best LLMs for Financial Analysis & Trading Decisions
Read business/financial docs deeply enough to make forward-looking judgments about opportunity and risk.
Read business/financial docs deeply enough to make forward-looking judgments about opportunity and risk.
Task-by-task breakdown
Onboarding Subject Analysis
Analyses a prospect's description to produce the full subject configuration: subject_name, subject_code, subject_type, industry, focus_areas, regions, suggested report cadence, and a recommended …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| MiniMax M3 ★ | 94% | RANKED | best value |
| Qwen 3.7 Plus | 94% | RANKED | 3.9x |
| Qwen 3.6 Flash | 92% | RANKED | 4.6x |
| Claude Sonnet 5 | 96% | HIGH | 7.4x |
| Gemini 3.5 Flash | 94% | RANKED | 7.7x |
| Grok 4.5 | 94% | RANKED | 9.1x |
| Meta Muse Spark 1.1 best | 100% | HIGH | 9.2x |
| GPT-5.6 Sol | 97% | MEDIUM | 14x |
Investment Panel Voting
Investment Panel Voting: a voter persona reviews all 10 panel members' analyses (7 voting + 3 advisory red-team) of the same subject and casts a structured vote — direction, expected % change, …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Qwen 3.6 Flash ★ | 92% | RANKED | best value |
| Qwen 3.7 Plus | 94% | RANKED | 1x |
| DeepSeek V4 Pro | 93% | HIGH | 1.1x |
| Qwen 3.6 Plus | 97% | HIGH | 1.8x |
| NVIDIA Nemotron-3 Ultra 550B | 90% | MEDIUM | 1.9x |
| Grok 4.5 | 97% | RANKED | 2.4x |
| GPT-5.6 Terra | 91% | MEDIUM | 2.6x |
| Meta Muse Spark 1.1 | 96% | MEDIUM | 3.7x |
| Kimi K2.6 | 95% | HIGH | 3.9x |
| Claude Sonnet 5 | 95% | RANKED | 4.4x |
| Claude Sonnet 4.6 best | 100% | RANKED | 5.1x |
| Claude Opus 4.8 | 98% | HIGH | 7.2x |
| GPT-5.5 | 95% | MEDIUM | 7.9x |
Trading Recommendation
Identifies market-moving catalysts and timing for an investment subject — categorised by timeframe and probability, with estimated magnitude (positive or negative). Distinguishes scheduled events from …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| GPT-5.4 Nano ★ | 91% | HIGH | best value |
| Grok 4.5 | 96% | RANKED | 1.8x |
| Qwen 3.6 Plus | 92% | MEDIUM | 2.5x |
| Gemini 3.5 Flash best | 100% | RANKED | 2.7x |
| Kimi K2.6 | 99% | RANKED | 6.4x |
| Claude Haiku 4.5 | 92% | MEDIUM | 7.9x |
| Claude Sonnet 4.6 | 99% | HIGH | 19x |
| GPT-5.5 | 99% | RANKED | 30x |
SEC Filing Analysis
Analyses a company's SEC filings (post-IPO) from a long-term investor's perspective — fundamental health, strategic positioning, long-term risks. Meticulous, objective, data-driven; returns structured …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| MiniMax M3 ★ best | 100% | RANKED | best value |
| Gemini 3.5 Flash | 91% | RANKED | 6x |
| Grok 4.5 | 91% | RANKED | 10x |
| GPT-5.5 | 92% | RANKED | 21x |
SEC S-1 Chunk Analysis
Per-section analysis of an S-1 / S-1/A registration statement for a long-term investor: business model, financial metrics, risk factors, strategic direction, market opportunity. Grounded only in the …
| Model | Quality (% of best) | Confidence | Overpay |
|---|---|---|---|
| Qwen 3.5 Flash ★ | 92% | RANKED | best value |
| GPT-5.4 Nano | 92% | RANKED | 1.5x |
| MiniMax M3 | 98% | RANKED | 1.8x |
| GPT-5.6 Luna | 96% | HIGH | 2.5x |
| Qwen 3.7 Plus | 93% | RANKED | 3.1x |
| Qwen 3.6 Flash | 92% | RANKED | 3.6x |
| Qwen 3.6 Plus | 94% | RANKED | 5x |
| Gemini 3.5 Flash | 97% | RANKED | 6x |
| GPT-5.6 Terra | 97% | HIGH | 6.5x |
| Grok 4.5 | 98% | RANKED | 9.5x |
| Meta Muse Spark 1.1 | 100% | HIGH | 10x |
| Claude Sonnet 5 | 96% | RANKED | 11x |
| Kimi K2.6 | 96% | RANKED | 12x |
| GPT-5.6 Sol | 99% | HIGH | 14x |
| Claude Opus 4.8 best | 100% | HIGH | 17x |
| GPT-5.5 | 98% | RANKED | 28x |
Confidence — how sure we are about the quality score (more judgments + more agreement = higher confidence): RANKED many independent judges scored this model's outputs and their agreement is very high (most confident) — HIGH many judges have scored it and they mostly agree (well-pinned) — MEDIUM enough judges have weighed in to publish, but they disagree more than we'd like (treat with a small grain of salt). LOW-confidence cells are hidden everywhere on the site. See the methodology for the exact thresholds.