Glossary
60 terms
- 1K agent tasks
- 200K tokens in, 20K out, each.
- 1K chats
- 2K tokens in, 500 out, each.
- 1K documents
- 20K tokens in, 1K out, each.
- Above fit
- Far above this model's index.
- Agentic
- Multi-step tasks with tools.
- API ID
- The model ID on OpenRouter.
- Batch output / 1M
- Batch price per 1M output tokens.
- Benchmark
- The benchmark name.
- Blended
- 3 parts input, 1 output.
- Budget
- Top-25% index, best per paid dollar.
- Cache read / 1M
- US dollars per 1M cached tokens.
- Capabilities
- Index per category, out of 100.
- Category
- The capability the benchmark measures.
- Cheapest API
- Lowest output price per 1M.
- Coding
- Code writing and repair.
- Context
- Maximum tokens in one request.
- Cost
- US dollars to run one task.
- Coverage
- Index weight that has results.
- Elo
- LMArena rating and rank.
- Fastest
- Fastest with top-25% index.
- First token
- Seconds to the first output token.
- Gap
- Index points below the top.
- In index
- Results count toward the index.
- Index
- All categories, out of 100.
- Input / 1M
- US dollars per 1M input tokens.
- Instruction
- Tasks with strict prompt rules.
- JSON output
- Output follows a given JSON schema.
- Knowledge
- Facts and expert knowledge.
- Knowledge cutoff
- Last date of training data.
- Lab
- The company that makes the model.
- Latest
- Release of the newest model.
- Math
- Math problems.
- Max output
- Maximum output tokens per request.
- Model
- The model name.
- Models
- Models with a result.
- Models
- Models from this lab.
- Multilingual
- Tasks in many languages.
- Multimodal
- Tasks with images and text.
- Newest
- Date of the newest result.
- Output / 1M
- US dollars per 1M output tokens.
- Place
- Place among results on this benchmark.
- Ranked
- Models with an index.
- Reasoning
- Logic problems and puzzles.
- Released
- Month of release.
- Result
- The score from the source.
- Retires
- Date OpenRouter removes the model.
- Smartest
- Highest index.
- Smartest open
- Highest index with open weights.
- Speed
- Output tokens per second.
- State
- Enough results for an index.
- Time
- Time to run one task.
- Tools
- Tool calls on OpenRouter.
- Top model
- Model with the highest index.
- Top result
- Best result on this benchmark.
- Top-3 spread
- Points from first to third.
- Type
- Maker, cloud, host or router.
- Updated
- Date the source changed the results.
- Weights
- Link to the published weights.
- Where to run
- APIs that serve this model.
- Year
- Year of release.