Ranking desk
AI Leaderboard Desk
This page separates three different questions. Frontier now answers what is current. Reliability Floor asks whether everyday basics are covered. Evaluated composite ranks models inside the benchmark-backed scored set.
Important
The old “composite leaderboard” confusion came from treating one scored table as the answer to everything. The frontier lane keeps newer launches visible immediately, while the evaluated composite only ranks models once there is enough public evidence to score them with some confidence.
| Model | State | Price |
|---|---|---|
| Claude Opus 5 (batch) Anthropic / 1.0M context | tracking tracking | $5.00 / $25.00 |
| GPT-5.6 Sol (batch) OpenAI / 1.1M context | tracking tracking | $5.00 / $30.00 |
| Claude Fable 5 (batch) Anthropic / 1.0M context | tracking tracking | $10.00 / $50.00 |
| Gemini 3.6 Flash (batch) Google / 1.0M context | tracking tracking | $1.50 / $7.50 |
| Claude Sonnet 5 (batch) Anthropic / 1.0M context | tracking tracking | $2.00 / $10.00 |
| GPT-5.6 Terra (batch) OpenAI / 1.1M context | tracking tracking | $2.50 / $15.00 |
| GPT-5.5 (batch) OpenAI / 1.1M context | tracking tracking | $5.00 / $30.00 |
| Claude Opus 4.8 (batch) Anthropic / 1.0M context | tracking tracking | $5.00 / $25.00 |
| Gemini 3.5 Flash (batch) Google / 1.0M context | tracking tracking | $1.50 / $9.00 |
| GPT-5.4 OpenAI / 1.1M context | tracking tracking | $2.50 / $15.00 |
| Grok 4.20 xAI / 2.0M context | tracking tracking | $1.25 / $2.50 |
| Qwen 3.6 Plus Alibaba / 1.0M context | tracking tracking | $0.33 / $1.95 |
| Claude Opus 5 Fast Anthropic / 0K context | tracking tracking | $10.00 / $50.00 |
| Gemini 3.5 Flash Lite (batch) Google / 1.0M context | tracking tracking | $0.30 / $2.50 |
| GPT-5.6 Luna (batch) OpenAI / 1.1M context | tracking tracking | $1.00 / $6.00 |
| Claude Opus 4.8 Fast Anthropic / 0K context | tracking tracking | $10.00 / $50.00 |
| GPT-5.5 Pro (batch) OpenAI / 1.1M context | tracking tracking | $30.00 / $180.00 |
| Claude Opus 4.6 (batch) Anthropic / 1.0M context | partial active | $5.00 / $25.00 |
The frontier lane is intentionally not a synthetic score. It keeps the current flagship and launch watchlist visible even when benchmark coverage is still catching up.
Evaluated composite uses only current, row-level source-backed benchmark results. Models need at least two distinct rankable benchmarks; coverage and evidence-count penalties reduce thin composites. The value tab divides that benchmark composite by current blended token price. The untraceable legacy quality layer is excluded.
Explore the field
Search and filter the full ranked dataset
The top tables are quick reads. This table is for actual decision-making when you want a specific lab, use case, release window, or pricing posture.
| Model | Composite | Coverage | Best for | Price |
|---|---|---|---|---|
| GPT-5.2 OpenAI / 400K context | 52.1 | 16% | Coding API | $0.88 / $7.00 |
| Gemini 3 Pro Google / 1.0M context | 51.4 | 9% | Reasoning API | $2.00 / $12.00 |
| Claude Opus 4.5 (batch) Anthropic / 200K context | 51.4 | 9% | Reasoning API | $2.50 / $12.50 |
| Claude Sonnet 4.5 (batch) Anthropic / 1.0M context | 49.1 | 9% | Reasoning API | $1.50 / $7.50 |