--- license: cc-by-4.0 tags: - llm - benchmarks - ai-evaluation - model-comparison - features - rate-limits size_categories: - n<1K --- # LLM Benchmark & Feature Matrix 2026 **Which LLM is best at what?** This dataset maps capabilities, performance, and limits of 22 major models. Unlike pricing datasets, this focuses on **what models can do** — not just what they cost. ## Files | File | Description | |------|-------------| | `llm-benchmarks-2026.csv` | MMLU, HumanEval, MATH, Arena ELO, coding/reasoning/multilingual rankings, tier (S+ to B) | | `llm-features-2026.csv` | 15 binary capabilities: vision, function calling, JSON mode, fine-tuning, tool use, web search, embeddings... | | `llm-rate-limits-2026.csv` | Free tier availability, RPM/TPM limits, batch discounts, cached input discounts | ## Models Covered 22 models from OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek, xAI, Cohere ## Use Cases - **Model selection** — Find models that support your required features (e.g., vision + function calling + fine-tuning) - **Performance comparison** — Which model scores highest on coding vs reasoning vs multilingual? - **Rate limit planning** — Can you stay within free tier? What are the paid RPM limits? - **Tier analysis** — S+ tier models vs A tier — is the premium worth it? ## Key Insights - Only **Google Gemini** supports all 15 features (vision, search, embeddings, fine-tuning, code execution) - **DeepSeek** offers 90% cached input discount — massive savings for repetitive workloads - **Groq** has highest free tier RPM (30) with lowest latency - S+ tier models (o1, Claude Opus 4, Gemini 2.5 Pro, DeepSeek R1) all score >89 MMLU ## Related - [LLM Rankings](https://comparedge.com/best/llm) — Live leaderboard - [AI Tools Pricing Dataset](https://huggingface.co/datasets/ComparEdge/ai-tools-pricing-2026) — 104 AI tools - [Kaggle: Full Pricing Data](https://www.kaggle.com/datasets/comparedge/llm-api-pricing-comparison-2026) - [GitHub Open Data](https://github.com/comparedge/awesome-saas-comparison-data) ## License CC BY 4.0 — [ComparEdge](https://comparedge.com)