Intelligence
Highest available normalized scores.
ETHEN RESEARCH / MODEL INTELLIGENCE
The working map of today's AI models.
A research portal for comparing capability, speed, latency, price, context, and openness across the normalized model catalog.
Highlights
Start with the signal that matters.
Compact rankings give the catalog a readable first pass before the deeper benchmark sections.
Model recommender
Start with a routing question and compare the signals that matter.
Open path →Coding and agentic paths
Review reasoning-backed routes and benchmark discovery.
Open path →Provider comparison
See model availability and provider-level catalog coverage.
Open path →Updates and methodology
Read how normalized records, sources, and FAQ disclosures are handled.
Open path →Intelligence overview
Capability, ranked across the catalog.
The overall intelligence view is built from the normalized intelligence records. Coding, agentic, and reasoning discovery paths remain available through the benchmark and leaderboard routes where distinct fields exist.
Overall intelligence
Top 24 records · higher is better
Intelligence segmentation
Open-weight and proprietary availability in the current catalog; no openness score is inferred.
Reasoning model distribution
Reasoning metadata is shown as provided by normalized model records.
Intelligence over time
Release dates give the scores a timeline.
This view plots the available normalized intelligence score against each model's recorded release date. It is a catalog trend, not a claim about benchmark history.
Intelligence by release date
Available model records with both release dates and intelligence scores.
Coding and agentic performance
Use the closest supported routing signal.
Distinct coding-agent or tool-use scores are not present in the directory schema. The reasoning distribution and intelligence leaderboard provide the closest supported discovery path without inventing a benchmark.
Reasoning-capable models
Models whose normalized records identify reasoning support.
Benchmark discovery
Open the intelligence and benchmark routes to inspect available coding, reasoning, and agentic dimensions from the canonical navigation.
Open path →Capability indices
Explore supported dimensions.
The current catalog exposes broad model metadata rather than a universal domain-score matrix. These views keep the available dimensions explicit.
Output tokens and context
Capacity changes what a model can hold.
Context-window values come from technical specifications in the normalized model records. Output-token capacity is not a separate directory field, so it is not fabricated here.
Largest context windows
Token capacity, shown in the repository's normalized units.
Context by provider
Provider coverage is shown through the models with context metadata.
Price and cost
Price efficiency without invented task economics.
Input and output prices are displayed in USD per 1M tokens. The repository does not provide a defensible universal task-cost formula, so this page uses price efficiency wording instead.
Lowest output prices
USD per 1M output tokens, lower is less expensive.
Intelligence and provider price paths
Use the full catalog to compare intelligence scores with exact input and output pricing.
Speed and latency
Responsiveness is a routing decision.
Output speed and latency are separate signals. The catalog preserves both so a fast generator is not confused with a low time-to-first-token model.
Output speed ranking
Output tokens per second, higher is faster.
Latency ranking
Available latency records, lower is faster.
API provider performance
Compare provider coverage from the catalog.
Provider-level benchmark measurements are not separately normalized in the directory. These views use truthful model availability and supported model-level signals by provider.
Models available by provider
Count of normalized models in the provider catalog.
Provider intelligence coverage
Highest available intelligence records grouped by provider.
Complete catalog
Search and compare every model.
Server-side search, sort, and pagination — the full library is not shipped to the browser.
550 models · catalog 2026-07-12 · 1ms
| Claude Fable 5 (with fallback) | Anthropic | 60 | 70.7 | 149.84s | $50.00 | 1M |
| Claude Opus 4.8 (max) | Anthropic | 56 | 64.5 | 37.32s | $50.00/M | 1M |
| GPT-5.5 (xhigh) | OpenAI | 55 | 88 | 94.38s | $20.00/m | 920k |
| Claude Opus 4.7 (max) | Anthropic | 54 | 56.4 | 17.52s | $25.00 | 1M |
| Claude Sonnet 5 (max) | Anthropic | 53 | 83.7 | 187.93s | $10.00/M | 1M |
| GPT-5.5 (high) | OpenAI | 53 | 83.1 | 26.89s | $30.00 | 920k |
| GLM-5.2 (max) | Z AI | 51 | 217.6 | 1.4s | $3.0 | 1M |
| GPT-5.4 (xhigh) | OpenAI | 51 | 175.3 | 113.48s | $1 | 1.1M |
| Gemini 3.5 Flash | 50 | 192.2 | 28.25s | $9.00 | 1M | |
| GPT-5.5 (medium) | OpenAI | 50 | 72.7 | 9.23s | $30.00 | 920k |
| Claude Sonnet 4.6 (max) | Anthropic | 47 | 71 | 137.72s | $15.00 | 1M |
| Gemini 3.1 Pro Preview | 46 | 151.6 | 27.42s | $12.00/M | 1M | |
| Qwen3.7 Max | Alibaba | 46 | 206.4 | 2.57s | $7.50 | 3.7M |
| Gemini 3.5 Flash (medium) | 45 | 191.4 | 19.51s | $9.00 | 1M | |
| Claude Opus 4.6 (max) | Anthropic | 44 | 43.7 | 23.27s | $25.00 | 1M |
| DeepSeek V4 Pro (max) | DeepSeek | 44 | 72 | 1.73s | $0.87 | 1M |
| GPT-5.3 Codex (xhigh) | OpenAI | 44 | 99 | 98.73s | $0.17/M | 400k |
| Kimi K2.6 | Kimi | 44 | 76.2 | 2.37s | $4 | 260k |
| MiniMax-M3 | MiniMax | 44 | 99 | 1.81s | $120/M | 1M |
| Claude Opus 4.7 (Non-reasoning, high) | Anthropic | 43 | 50.5 | 1.52s | $25.00 | 1M |
| GPT-5.5 (low) | OpenAI | 43 | 81.7 | 1.79s | $30.00 | 920k |
| Muse Spark | Meta | 43 | — | — | $0.00 | 260k |
| Claude Sonnet 5 (Non-reasoning) | Anthropic | 42 | — | — | $10.00 | 1M |
| GPT-5.2 (xhigh) | OpenAI | 42 | 84.4 | 202.88s | $0.17/M | 400k |
| Kimi K2.7 Code | Kimi | 42 | 51.4 | 2.91s | $0.15/M | 260k |
| MiMo-V2.5-Pro | Xiaomi | 42 | 52.5 | 2.77s | $0.87/M | 1M |
| Claude Opus 4.5 | Anthropic | 41 | 59.7 | 13.31s | $25.00 | 200k |
| DeepSeek V4 Pro (high) | DeepSeek | 41 | 64.8 | 1.66s | $0.87 | 1M |
| Nex-N2-Pro | Nex AGI | 41 | 92.3 | 1.81s | $2.50 | 260k |
| DeepSeek V4 Flash (max) | DeepSeek | 40 | 113 | 1.26s | $00/M | 1M |
| Gemini 3 Pro Preview (high) | 40 | — | 48s | $12.00 | 1M | |
| GLM-5 | Z AI | 40 | 52.5 | 1.64s | $2 | 200k |
| GLM-5.1 | Z AI | 40 | 79.6 | 1.46s | $026/M | 200k |
| GPT-5.2 Codex (xhigh) | OpenAI | 40 | 161 | 45.58s | $0.17/M | 400k |
| GPT-5.4 mini (xhigh) | OpenAI | 40 | 179.3 | 12.31s | $4.50 | 400k |
| Grok Build 0.1 0616 | xAI | 40 | 69.1 | 0.54s | $200/m | 260k |
| MiMo-V2-Pro | Xiaomi | 40 | — | — | $3.00/M | 1M |
| MiMo-V2.5 | Xiaomi | 40 | 84.9 | 2.83s | $0.0/M | 1M |
| Qwen3.6 Max Preview | Alibaba | 40 | 45.1 | 3.38s | $7.80 | 260k |
| Qwen3.6 Plus | Alibaba | 40 | 51.8 | 2.79s | $3.00/M | 1M |
| GPT-5.4 (low) | OpenAI | 39 | 129.5 | 1.88s | $15.00 | 1.1M |
| Qwen3.7 Plus | Alibaba | 39 | 50.5 | 2.91s | $1.16 | 1M |
| Claude Opus 4.6 (high) | Anthropic | 38 | 42.6 | 2.62s | $25.00 | 1M |
| Gemini 3 Flash | 38 | 176 | 7.99s | $3.00 | 1M | |
| GLM-5-Turbo | Z AI | 38 | — | 47s | $024/M | 200k |
| GPT-5.2 (medium) | OpenAI | 38 | — | — | $14.00 | 400k |
| GPT-5.4 nano (xhigh) | OpenAI | 38 | 152.9 | 6.45s | $0.02/M | 400k |
| Grok 4.3 (high) | xAI | 38 | 164.2 | 21.56s | $02/M | 1M |
| Kimi K2.5 | Kimi | 38 | 43.1 | 3s | $3.00/m | 260k |
| MiniMax-M2.7 | MiniMax | 38 | 51.1 | 2.01s | $1.20 | 200k |
Page 1 of 11