Choose the right AI model for the work—not the hype.
Compare capability, cost, context, and deployment tradeoffs with source-aware guidance built for people shipping real AI systems.
models tracked
providers reviewed
latest source review
See sources and caveats →
What are you trying to do?
A benchmark only matters when it resembles your constraints. Choose a decision path before choosing a winner.
Ship reliable code changes
Prioritize coding quality, tool use, and long-context repo work.
Explore the guide 02ReasoningSolve difficult, constrained work
Compare reasoning depth without hiding latency and price tradeoffs.
Explore the guide 03High volumeControl inference cost
Find the quality floor that meets your workload at production scale.
Explore the guide 04Private deploymentRun open weights locally
Balance privacy, hardware requirements, and operational overhead.
Explore the guideThe current leaderboard
One editorial composite for orientation, with the underlying dimensions, price, and source trail kept visible.
| Rank | Model | Editorial fit | Coding | Reasoning | Tool use | Input / output |
|---|---|---|---|---|---|---|
| 01 | Claude Fable 5Anthropic | 9.9 | 9.9 | 9.9 | 9.8 | $10 / $50 |
| 02 | GPT-5.5OpenAI | 9.8 | 9.8 | 9.8 | 9.7 | $5 / $30 |
| 03 | Claude Opus 4.8Anthropic | 9.8 | 9.8 | 9.8 | 9.6 | $5 / $25 |
| 04 | GPT-5.4OpenAI | 9.7 | 9.8 | 9.5 | 9.7 | $2.5 / $15 |
| 05 | Claude Sonnet 5Anthropic | 9.5 | 9.6 | 9.5 | 9.3 | $3 / $15 |
| 06 | GPT-5.2-CodexOpenAI | 9.5 | 9.7 | 9.3 | 9.4 | $1.75 / $14 |
Scores are editorial decision aids, not universal benchmark claims.
View all models →Evidence before verdicts.
Model choice is multidimensional. We separate source-backed specifications from editorial interpretation, date every review, and show caveats where provider tiers or deployment choices change the answer.
Read the methodology →Verify the facts
Pricing, context, availability, and release details link back to provider or benchmark sources.
Expose the weights
Composite scores state what they reward so you can decide whether the ranking fits your job.
Name the tradeoffs
Every recommendation should include a reason to choose it and a reason not to.
Recent evaluations
Daily Model Eval Scorecard — 2026-04-11
Head-to-head results across coding, reasoning, and tool-use tasks. Today: Gemini 3.1 Pro Preview, GPT-5.4 XHigh, Grok 4.20, and Gemma 4.
Read analysis →Daily Model Eval Scorecard — 2026-04-10
Head-to-head results across coding, reasoning, and tool-use tasks. Today: Gemini 3.1 Pro Preview, GPT-5.4 XHigh, Muse Spark, and GLM-5.1.
Read analysis →Daily Model Eval Scorecard — 2026-04-07
Head-to-head results across coding, reasoning, and tool-use tasks. Today: Gemini 3.1 Pro Preview, GPT-5.4 XHigh, Gemma 4, and Qwen 3.5.
Read analysis →