Showing 10 of 386 models for coding
Explore models and compare hosted costs.
API prices are in USD per 1 million tokens. Input is what you send; output is what the model generates. Both are billed separately. These are hosted API rates, not ChatGPT/Claude subscriptions or local running costs.
Task fit uses sourced evaluations where a tested configuration is matched. Writing uses a relative creative-writing percentile; support and agents use banking tool-use success. Research-system scores are shown separately. “—” means unknown, not zero. See task evidence and limitations →
| # | Add | Model | Fit for Coding | Size | Runs | Speed | API price / 1M tokens | Context | |
|---|---|---|---|---|---|---|---|---|---|
| 1 | anthropic · Weights unverified | Fair | — | ☁ HostedCloud available | — | Input $4 Output $20OpenRouter | 1,000,000 | ||
| 2 | openai · Weights unverified | Fair | — | ☁ HostedCloud available | — | Input $10 Output $50OpenRouter | 1,050,000 | ||
| 3 | anthropic · Weights unverified | Fair | — | ☁ HostedCloud available | — | Input $10 Output $50OpenRouter | 1,000,000 | ||
| 4 | anthropic · Weights unverified | Fair | — | ☁ HostedCloud available | — | Input $5 Output $25OpenRouter | 1,000,000 | ||
| 5 | openai · Weights unverified | Weak | — | ☁ HostedCloud available | — | Input $2 Output $10OpenRouter | 1,050,000 | ||
| 6 | openai · Weights unverified | Weak | — | ☁ HostedCloud available | — | Input $2 Output $10OpenRouter | 1,050,000 | ||
| 7 | xiaomi · Open weights | Weak | — | ↓ Open weightsCloud available | — | Input $0.435 Output $0.87OpenRouter | 1,048,576 | ||
| 8 | StepFun · Weights unverified | Weak | — | ☁ HostedHosted availability — | — | Input — Output —No hosted quote | — | ||
| 9 | meta · Weights unverified | Weak | — | ☁ HostedCloud available | — | Input $1.25 Output $4.25OpenRouter | 1,048,576 | ||
| 10 | deepseek · Open weights | Weak | — | ↓ Open weightsCloud available | — | Input $0.14 Output $0.42OpenRouter | 1,048,576 |
Newest means date added to the source catalogue, not a verified release date or quality ranking.
Local fit is a memory estimate, not a measured benchmark. Open a row for sources and assumptions.