The newest model in the AIM LLM Index is GPT-6.2 Astra.
AIM LLM Index
v1.11One transparent index combining 5 public benchmarks with AIMultiple's 4 agentic, RAG and enterprise-reasoning evaluations. See AIM LLM Index Methodology for further details.
| Name | Series | Cost per task ($, log scale) | LLM Index |
|---|---|---|---|
| Gemini 4 Carbon | 0.0402 | 85 | |
| Gemini 4 Argon | 0.0459 | 73 | |
| Gemini 3.8 Flash | 0.013 | 58 | |
| GPT-6.2 Astra | OpenAI | 0.0794 | 83 |
| GPT-6 Astra | OpenAI | 0.455 | 73 |
| GPT-6.1 Sol | OpenAI | 0.0225 | 73 |
| GPT-6.1 Terra | OpenAI | 0.0132 | 69 |
| GPT-6 Sol | OpenAI | 0.0314 | 69 |
| GPT-6 Luna | OpenAI | 0.0017 | 57 |
| Claude Fable 5.2 | Anthropic | 0.1452 | 77 |
| Claude Opus 5.5 | Anthropic | 0.1148 | 76 |
| Claude Fable 5.1 | Anthropic | 0.3895 | 72 |
| Claude Sonnet 5.5 | Anthropic | 0.0386 | 72 |
| Claude Haiku 5.5 | Anthropic | 0.0025 | 56 |
| Muse Wonder | Meta | 0.1107 | 68 |
| Muse Spark 1.3 | Meta | 0.04 | 61 |
| Muse Spark Open | Meta | 0.0493 | 53 |
| Kimi K3.1 Pro | Moonshot AI | 0.0438 | 67 |
| MiMo-V2.6-Pro | Xiaomi | 0.008 | 58 |
| MiniMax M3.1 Flash | MiniMax | 0.0045 | 56 |
| Grok 4.7 | X | 0.0659 | 56 |
| Qwen3.8 Flash | Alibaba Cloud | 0.005 | 56 |
| Qwen3.8 Max (0902) | Alibaba Cloud | 0.0526 | 55 |
| DeepSeek V4.1 Flash 1017 | DeepSeek | 0.0023 | 56 |
| DeepSeek V4.1 Flash | DeepSeek | 0.003 | 52 |
| GLM 5.3 Flash | Z AI | 0.0014 | 53 |
| Name | AIM LLM Index |
|---|---|
| Gemini 4 Carbon | 85 |
| GPT-6.2 Astra | 83 |
| Claude Fable 5.2 | 77 |
| Claude Opus 5.5 | 76 |
| GPT-6 Astra | 73 |
| Gemini 4 Argon | 73 |
| GPT-6.1 Sol | 73 |
| Claude Fable 5.1 | 72 |
| Claude Sonnet 5.5 | 72 |
| GPT-6.1 Terra | 69 |
| GPT-6 Sol | 69 |
| Muse Wonder | 68 |
| Kimi K3.1 Pro | 67 |
| MiMo-V2.6-Pro | 58 |
| GPT-6 Luna | 57 |
| MiniMax M3.1 Flash | 56 |
| Grok 4.7 | 56 |
| DeepSeek V4.1 Flash 1017 | 56 |
| Claude Haiku 5.5 | 56 |
| Muse Spark Open | 53 |
Leaderboard
The highest-scoring models across all benchmarks.
# | Model | Index | AIM-Agentic-RAG | AIM-AI-Bias | AIM-FinanceReasoning | AIM-Text-to-SQL | CritPt | FrontierMath | MMMU-Pro | SciCode | Terminal-Bench 4.0 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Gemini 4 Carbon Google | 85 | 92 | - | 90 | 78 | - | 95 | 89 | - | 65 |
| 2 | GPT-6.2 Astra OpenAI | 83 | 86 | 88 | 88 | 68 | - | 100 | 89 | - | 64 |
| 3 | Claude Fable 5.2 Anthropic | 77 | 97 | 74 | 92 | 86 | 36 | 100 | - | 68 | 62 |
| 4 | Claude Opus 5.5 Anthropic | 76 | 90 | 78 | 92 | 88 | 32 | 95 | 88 | 67 | 60 |
| 5 | GPT-6 Astra OpenAI | 73 | 84 | 91 | 87 | 66 | 32 | 98 | 87 | 56 | 60 |
| 6 | Gemini 4 Argon Google | 73 | 94 | 84 | 91 | 79 | 27 | 78 | 88 | 62 | 57 |
| 7 | GPT-6.1 Sol OpenAI | 73 | 81 | 93 | 90 | 65 | 32 | 100 | 86 | 56 | 56 |
| 8 | Claude Fable 5.1 Anthropic | 72 | 84 | 85 | 90 | 67 | 31 | 88 | 86 | 63 | 55 |
| 9 | Claude Sonnet 5.5 Anthropic | 72 | 85 | 74 | 89 | 74 | 31 | 80 | 86 | 61 | 64 |
| 10 | GPT-6.1 Terra OpenAI | 69 | 84 | 82 | 87 | 57 | 28 | 93 | 85 | 55 | 56 |
| Name | Series | Model release date | LLM Index | Release |
|---|---|---|---|---|
| Claude Opus 4.8 | Anthropic | 2026-05-27 | 59 | May 27, 2026 |
| Claude Fable 5 | Anthropic | 2026-06-09 | 68 | Jun 9, 2026 |
| Claude Fable 5.1 | Anthropic | 2026-09-01 | 72 | Sep 1, 2026 |
| Claude Opus 5.5 | Anthropic | 2026-09-22 | 76 | Sep 22, 2026 |
| Claude Fable 5.2 | Anthropic | 2026-11-09 | 77 | Nov 9, 2026 |
| Gemini 3.6 Flash | 2026-07-21 | 53 | Jul 21, 2026 | |
| Gemini 3.7 Flash | 2026-08-13 | 60 | Aug 13, 2026 | |
| Gemini 4 Argon | 2026-10-13 | 73 | Oct 13, 2026 | |
| Gemini 4 Carbon | 2026-11-10 | 85 | Nov 10, 2026 | |
| DeepSeek V4 Flash 0731 | DeepSeek | 2026-07-31 | 48 | Jul 31, 2026 |
| DeepSeek V4 Pro 0813 | DeepSeek | 2026-08-12 | 49 | Aug 12, 2026 |
| DeepSeek V4.1 Flash | DeepSeek | 2026-09-10 | 52 | Sep 10, 2026 |
| DeepSeek V4.1 Flash 1017 | DeepSeek | 2026-10-17 | 56 | Oct 17, 2026 |
| Muse Spark 1.1 | Meta | 2026-07-16 | 55 | Jul 16, 2026 |
| Muse Spark 1.2 | Meta | 2026-08-05 | 57 | Aug 5, 2026 |
| Muse Spark 1.3 | Meta | 2026-09-02 | 61 | Sep 2, 2026 |
| Muse Wonder | Meta | 2026-10-12 | 68 | Oct 12, 2026 |
| GPT-5.6 Sol | OpenAI | 2026-07-09 | 67 | Jul 9, 2026 |
| GPT-6 Astra | OpenAI | 2026-09-04 | 73 | Sep 4, 2026 |
| GPT-6.2 Astra | OpenAI | 2026-11-17 | 83 | Nov 17, 2026 |
| Kimi K3 | Moonshot AI | 2026-07-16 | 57 | Jul 16, 2026 |
| Kimi K3.1 Pro | Moonshot AI | 2026-10-23 | 67 | Oct 23, 2026 |
| MiniMax M3 | MiniMax | 2026-05-31 | 40 | May 31, 2026 |
| MiniMax M3.1 Flash | MiniMax | 2026-10-19 | 56 | Oct 19, 2026 |
| Grok 4.5 | X | 2026-07-08 | 54 | Jul 8, 2026 |
| Grok 4.6 | X | 2026-08-12 | 59 | Aug 12, 2026 |
| Qwen3.7 Plus | Alibaba Cloud | 2026-06-03 | 43 | Jun 3, 2026 |
| Qwen3.8 Max | Alibaba Cloud | 2026-08-03 | 58 | Aug 3, 2026 |
| GLM 5.2 | Z AI | 2026-06-16 | 49 | Jun 16, 2026 |
| GLM 5.3 | Z AI | 2026-08-18 | 57 | Aug 18, 2026 |
| Model | Benchmark | Score |
|---|---|---|
| GPT-6.2 Astra | AIM-Agentic-RAG | 86 |
| GPT-6.2 Astra | AIM-FinanceReasoning | 88 |
| GPT-6.2 Astra | AIM-Text-to-SQL | 68 |
| GPT-6.2 Astra | Terminal-Bench 4.0 | 64 |
| GPT-6.2 Astra | AIM-AI-Bias | 88 |
| Claude Fable 5.2 | AIM-Agentic-RAG | 97 |
| Claude Fable 5.2 | AIM-FinanceReasoning | 92 |
| Claude Fable 5.2 | AIM-Text-to-SQL | 86 |
| Claude Fable 5.2 | Terminal-Bench 4.0 | 62 |
| Claude Fable 5.2 | AIM-AI-Bias | 74 |
| Claude Opus 5.5 | AIM-Agentic-RAG | 90 |
| Claude Opus 5.5 | AIM-FinanceReasoning | 92 |
| Claude Opus 5.5 | AIM-Text-to-SQL | 88 |
| Claude Opus 5.5 | Terminal-Bench 4.0 | 60 |
| Claude Opus 5.5 | AIM-AI-Bias | 78 |
| GPT-6 Astra | AIM-Agentic-RAG | 84 |
| GPT-6 Astra | AIM-FinanceReasoning | 87 |
| GPT-6 Astra | AIM-Text-to-SQL | 66 |
| GPT-6 Astra | Terminal-Bench 4.0 | 60 |
| GPT-6 Astra | AIM-AI-Bias | 91 |
| Gemini 4 Argon | AIM-Agentic-RAG | 94 |
| Gemini 4 Argon | AIM-FinanceReasoning | 91 |
| Gemini 4 Argon | AIM-Text-to-SQL | 79 |
| Gemini 4 Argon | Terminal-Bench 4.0 | 57 |
| Gemini 4 Argon | AIM-AI-Bias | 84 |
| Name | Series | Blended cost ($ / 1M tokens, log scale) | LLM Index | Release |
|---|---|---|---|---|
| Gemini 4 Carbon | 2 | 85 | 2026-11-10 | |
| Gemini 4 Argon | 4 | 73 | 2026-10-13 | |
| Gemini 3.8 Flash | 0.75 | 58 | 2026-09-02 | |
| GPT-6.2 Astra | OpenAI | 16 | 83 | 2026-11-17 |
| GPT-6 Astra | OpenAI | 20 | 73 | 2026-09-04 |
| GPT-6.1 Sol | OpenAI | 4 | 73 | 2026-09-29 |
| GPT-6.1 Terra | OpenAI | 2 | 69 | 2026-10-26 |
| GPT-6 Sol | OpenAI | 4 | 69 | 2026-09-22 |
| GPT-6 Luna | OpenAI | 0.2 | 57 | 2026-09-22 |
| Claude Fable 5.2 | Anthropic | 16 | 77 | 2026-11-09 |
| Claude Opus 5.5 | Anthropic | 8 | 76 | 2026-09-22 |
| Claude Fable 5.1 | Anthropic | 20 | 72 | 2026-09-01 |
| Claude Sonnet 5.5 | Anthropic | 4 | 72 | 2026-09-28 |
| Claude Haiku 5.5 | Anthropic | 0.2 | 56 | 2026-10-07 |
| Muse Wonder | Meta | 8 | 68 | 2026-10-12 |
| Muse Spark 1.3 | Meta | 2 | 61 | 2026-09-02 |
| Muse Spark Open | Meta | 2 | 53 | 2026-10-15 |
| Kimi K3.1 Pro | Moonshot AI | 3.8 | 67 | 2026-10-23 |
| MiMo-V2.6-Pro | Xiaomi | 0.49 | 58 | 2026-09-21 |
| MiniMax M3.1 Flash | MiniMax | 0.24 | 56 | 2026-10-19 |
| Grok 4.7 | X | 3 | 56 | 2026-09-21 |
| Qwen3.8 Flash | Alibaba Cloud | 0.23 | 56 | 2026-08-26 |
| Qwen3.8 Max (0902) | Alibaba Cloud | 3 | 55 | 2026-09-03 |
| DeepSeek V4.1 Flash 1017 | DeepSeek | 0.15 | 56 | 2026-10-17 |
| DeepSeek V4.1 Flash | DeepSeek | 0.21 | 52 | 2026-09-10 |
| GLM 5.3 Flash | Z AI | 0.09 | 53 | 2026-08-26 |
Frequently asked questions
AIM LLM Index Methodology
A model's index score is the average of its results across the benchmarks below. Scores already on a 0-100 scale are used as-is; others are rescaled to 0-100. A model missing a benchmark has it excluded rather than scored as zero, and every benchmark carries equal weight. A model needs to be scored in at least 2 AIMultiple benchmarks to be ranked. MMMU-Pro results come from 2 separate sources.
Benchmarks We Used
Raw results of 2.5 thousand data points from the AIMultiple benchmarks this index is built from. Download them as a ZIP file containing 4 CSV files and a README.
Want the granular data behind it? Join Premium
Changelog
70 updates- v1.11
Added 119 models to the dataset: Claude 3 Haiku, Claude Opus 4.7 (max), Command A, DeepSeek R1 (Jan), DeepSeek R1 0528 Qwen3 8B and 114 more.
- v1.11
2 models entered the AIM LLM Index: GPT-6.2 Astra, GPT-6.2 Sol.
- v1.11
Added 2 models to the dataset: GPT-6.2 Astra, GPT-6.2 Sol.
- v1.11
AIM-AI-Bias was added to the index calculation. Version updated to v1.11.
- v1.10
2 models entered the AIM LLM Index: Gemini 4 Hydrogen, Qwen4 Max.
- v1.10
Added 3 models to the dataset: Gemini 4 Hydrogen, Hy4, Qwen4 Max.
- v1.10
Gemini 4 Carbon entered the AIM LLM Index.
- v1.10
Added Gemini 4 Carbon to the dataset.
- v1.10
2 models entered the AIM LLM Index: Claude Fable 5.2, Muse Ember.
- v1.10
Added 2 models to the dataset: Claude Fable 5.2, Muse Ember.
Explore LLM Use Cases, Analyses & Benchmarks
- LLMBenchmarkNov 10
AI Gateways for OpenAI: OpenRouter Alternatives
We benchmarked OpenRouter, SambaNova, TogetherAI, Groq, and AI/ML API across three indicators (first-token latency, total latency, and output-token count), with 300 tests using short prompts (approx. 18 tokens) and long prompts (approx. 203 tokens) for total latency. If you plan to use one of these AI gateways, you can: In this benchmark, we compared OpenRouter,…
- LLMOpen World EvaluationNov 9
LLM Orchestration: 21 Frameworks and Gateways
Optimizing LLM orchestration is key to improving performance while keeping resource use under control. To evaluate how different orchestration approaches perform in practice, we benchmarked: Discover selected LLM orchestration tools, including developer frameworks and enterprise gateways: LLM Orchestration involves managing and integrating multiple Large Language Models (LLMs) to perform complex tasks efficiently. It ensures interaction…
- LLMFeature ComparisonNov 5
LLM Pricing: Top 15+ Providers Compared
We track the launch prices of more than 160 LLMs, from GPT-3.5 Turbo in March 2023 to the latest releases, grouped by size class. We also compare current models’ API prices with their benchmark results and review the subscription plans of the major providers. Prices are per million tokens, blended three parts input to one…
- LLMInsightNov 5
LLM Market Share: Compare Usage & Adoption
We analyzed LLM market share by combining usage-based data and web visit estimates to show how demand for large language models is distributed across AI labs and AI applications. Read the methodology to see how we measured and calculated these results. The United States dominated web visits across all four months, consistently accounting for 90–95%.…
- LLMInsightNov 2
LLM Parameters: Types, Scaling, Fine-Tuning Guide
Parameter count is one of the most cited specifications of a large language model, but it does not determine how well a model performs or how much it costs to run. The chart below compares 31 models by their LLM Index score and their average cost per task across seven benchmarks. Four models form the…
- LLMInsightOct 28
The Future of Large Language Models
See the future of large language models by comparing approaches, such as self-training, fact-checking, and sparse expertise, that could address LLM limitations. LLMs access external sources during conversations instead of relying on training data. The model queries external databases, retrieves current information, and provides citations. Limitation: Still makes errors. Citations don’t guarantee accuracy; models sometimes…
- LLMInsightOct 26
50+ ChatGPT Use Cases with Real Life Examples
ChatGPT reached approximately 1 billion weekly active users in early 2026 roughly 10% of the world’s population. OpenAI surpassed $20 billion in annual revenue for 2025, confirmed by CFO Sarah Friar. The Anthropic Economic Index distinguishes two modes of use: augmentation, in which a human interacts with AI, and automation, in which AI completes tasks…
- LLMBenchmarkOct 26
HALC-Bench: LLM Hallucination on Long-Context Retrieval Benchmark
HALC-Bench (LLM Hallucination on Long-Context Retrieval Benchmark) measures a large language model’s resistance to fabricating evidence for a metric that does not exist in the target document by using 3 haystacks placed at the beginning, middle, and end of the model’s context window, with 204 questions. claude-fable-5 answered all 204 traps correctly at every haystack…