aster-3.5
Handles multi-step instructions, long context, and structured output. The safest default for customer-facing assistants.
All models share the same API surface. Start with aster-3.5, then move down in size when latency or cost matters more than depth.
| Model | Context | Input / 1M | Output / 1M | Median TTFT | Best for |
|---|---|---|---|---|---|
aster-3.5 |
128k | $0.80 | $2.40 | 38 ms | General chat, long documents, tool use |
aster-3.5-mini |
64k | $0.16 | $0.48 | 19 ms | Classification, extraction, high volume |
aster-code-2 |
96k | $0.55 | $1.65 | 31 ms | Code generation, review, refactoring |
aster-embed-1 |
8k | $0.04 | — | 9 ms | Search, clustering, retrieval |
Handles multi-step instructions, long context, and structured output. The safest default for customer-facing assistants.
Optimized for short prompts and high request volume. Good for routing, tagging, and lightweight agents.
Trained for repository context, diffs, and test generation. Supports JSON schema constraints.
Start with the sandbox tier and compare outputs on your own data.