The Aggregate: unified LLM rankings, daily
Best AI you can use today: Claude Opus 5.5.
We fold every public benchmark we can find into one daily ranking. Unified rankings for 2171 AI models in the fused corpus across 26518 public benchmarks in the current fit, aggregated from public leaderboards and refreshed daily. Rankings use an Item Response Theory (IRT) fit that maps every model onto one rating scale.
Top models by Aggregate Rating
| Rank | Model | Provider | Rating | Benchmarks |
|---|---|---|---|---|
| 1 | Claude Mythos 5.1 | Anthropic | 2019 ± 29 | 23 |
| 2 | Claude Mythos 5 | Anthropic | 2009 ± 13 | 105 |
| 3 | Claude Opus 5.5 | Anthropic | 1986 ± 5 | 713 |
| 4 | Claude Mythos Preview | Anthropic | 1984 ± 12 | 97 |
| 5 | GPT-6 Astra Pro | OpenAI | 1984 ± 21 | 20 |
| 6 | GPT-6 Astra | OpenAI | 1983 ± 4 | 1009 |
| 7 | GPT-6.1 Sol | OpenAI | 1969 ± 6 | 613 |
| 8 | Gemini 4 Argon | 1967 ± 7 | 429 | |
| 9 | Claude Fable 5.1 | Anthropic | 1958 ± 5 | 631 |
| 10 | Muse Wonder Full | Meta | 1939 ± 20 | 133 |
| 11 | Claude Sonnet 5.5 | Anthropic | 1934 ± 7 | 482 |
| 12 | Claude Opus 5 | Anthropic | 1931 ± 3 | 1298 |
| 13 | GPT-6.1 Pro Sol | OpenAI | 1924 ± 29 | 18 |
| 14 | GPT-6 Sol | OpenAI | 1910 ± 6 | 577 |
| 15 | Muse Wonder | Meta | 1908 ± 9 | 242 |
| 16 | Claude Fable 5 | Anthropic | 1907 ± 4 | 914 |
| 17 | GPT-5.5 Pro | OpenAI | 1899 ± 13 | 76 |
| 18 | GPT-5.6 Pro Sol | OpenAI | 1890 ± 14 | 62 |
| 19 | GPT-5.6 Sol | OpenAI | 1889 ± 3 | 1958 |
| 20 | DeepSeek V4.1 Flash (1017) | DeepSeek | 1874 ± 13 | 230 |
Which AI is worth paying for?
Plans ranked by the model work a dollar buys, counting your time.
| Plan | Value per dollar | Price | Best use of the allowance | Verdict |
|---|---|---|---|---|
| Claude Max 20x | 1.2× | USD 200/mo | Claude Opus 5.5, 29B tokens a month, 158 h of your time | Higher Claude usage plus a separate monthly API budget for custom apps and agents; credits require claiming and cannot extend interactive plan limits. |
| Claude Pro | 1.2× | USD 20/mo | Claude Opus 5.5, 2.9B tokens a month, 16 h of your time | Generous Opus allowance for regular chat and coding in one subscription. |
| Claude Max 5x | 1.2× | USD 100/mo | Claude Opus 5.5, 14B tokens a month, 77 h of your time | Higher Claude usage plus a separate monthly API budget for custom apps and agents; credits require claiming and cannot extend interactive plan limits. |
| ChatGPT Pro 200 | 0.92× | USD 200/mo | GPT-6 Astra, 1.9B tokens a month, 10 h of your time | More agent headroom without a five-hour limit, assessed on the reduced allowance for new subscribers. |
| ChatGPT Pro 100 | 0.90× | USD 100/mo | GPT-6 Astra, 867M tokens a month, 4.7 h of your time | More agent capacity without a five-hour limit, with the broader ChatGPT toolset. |
| GitHub Copilot Max | 0.78× | USD 100/mo | Claude Opus 5.5, 577M tokens a month, 3.1 h of your time | A sizeable published coding budget for people who will use the extra credits. |
| Abacus ChatLLM | 0.78× | USD 10/mo | Claude Opus 5.5, 58M tokens a month, 19 min of your time | Many models for little money; unclear caps and fallbacks weaken the deal. |
| ChatGPT Plus | 0.76× | USD 20/mo | GPT-6 Astra, 106M tokens a month, 34 min of your time | A strong all-round bundle; sustained agent work can outgrow its limits. |
Latest leaderboard changes
- Gemini 4 Argon took the lead on Kaggle Game Arena Four in a Row (923.33 vs 695.32 by GPT-6 Astra)
- Gemini 4 Argon took the lead on Kaggle Game Arena (Unified) (535.09 vs 492.31 by GPT-6 Astra)
- Claude Opus 5.5 (Adaptive Reasoning, Max Effort) took the lead on HANDBOOK.md Agents (52.3 vs 38.8 by Claude Fable 5.1 (Max))
- DeepSeek V4.1 Flash (1017) took the lead on LLM Stats (AutomationBench; v1.0.6 method cohort) (60.5 vs 48.8 by GLM-5.3 Flash)
- MiniMax-M3.1-Flash took the lead on LLM Stats (AutomationBench v1.0.6) (62.2 vs 53.1 by MiMo-V2.6-Pro)
- Gemini 4 Argon (High) took the lead on APEX v1 (81.1 vs 75.6 by Muse Spark 1.2 (xHigh))
- DeepSeek V4.1 Flash (1017) took the lead on LLM Stats (NL2Repo) (68.2 vs 64 by DeepSeek V4.1 Flash)
- DeepSeek V4.1 Flash (1017) took the lead on LLM Stats (BabyVision) (92.5 vs 89.6 by DeepSeek V4.1 Flash)
Explore: all models · benchmark catalog · capability trends · daily changes
Interactive version: theaggregate.ai/ · How It Works · Data refreshed daily, snapshot 2026-10-20.