Models and Capabilities
Discover models, context limits, capabilities, and pricing from the model catalog.
GET https://api.abhigya.ai/v1/models is the source of truth for which models are available and what each one supports. Read the catalog at runtime instead of hard-coding model IDs.
List models
The response is an OpenAI-compatible model list. Use a model's id as the model field in requests. If the key is invalid, the request returns 401.
Available models
Prices are per million tokens in Indian rupees. Read them from /v1/models with your API key.
| Model ID | Context | Input | Output | Cached input |
|---|---|---|---|---|
deepseek/deepseek-v4.1-flash | 1M | ₹20.00 | ₹60.00 | ₹0.60 |
moonshot/kimi-k3 | 1M | ₹285.00 | ₹1,425.00 | ₹28.50 |
zai/glm-5.3-flash | 1M | ₹15.00 | ₹50.00 | ₹3.00 |
zai/glm-5.3 | 1M | ₹90.00 | ₹400.00 | ₹20.00 |
qwen/qwen3.8-flash-next | 256K | ₹15.00 | ₹47.00 | ₹1.60 |
qwen/qwen3.8-27b | 256K | ₹20.00 | ₹250.00 | ₹5.00 |
qwen/qwen2.5-7b-instruct | 128K | ₹10.00 | ₹20.00 | — |
google/gemma-4-31b-it | 256K | ₹20.00 | ₹40.00 | — |
google/gemma-4-26b-a4b-it | 256K | ₹7.00 | ₹34.00 | — |
Limits
| Field | Meaning |
|---|---|
| Context window | Prompt tokens plus generated tokens. A request that does not fit returns 400 with code context_length_exceeded. |
| Output limit | Maximum generated tokens per request. A larger max_tokens is lowered to the model's cap. See Output length. |
Capabilities
All listed models support Chat Completions with streaming. Chat Completions also supports function tools, JSON mode, and JSON Schema structured outputs.
- Function calling with JSON Schema tool definitions.
- Streaming responses and incremental tool-call arguments.
- Structured outputs with
response_format. - Reasoning effort controls on reasoning models.
Feature support can vary by model. Check the model catalog before relying on tools, structured outputs, or reasoning in production.
Reasoning efforts
Reasoning models accept reasoning_effort of none, low, medium, high, or max. A model applies the nearest tier it supports. An unrecognized value on Chat Completions returns 400 with code unsupported_value.
Reasoning tokens count toward max_tokens and bill as output tokens. Give reasoning requests enough headroom. See Reasoning.
Pricing
Serverless is prepaid and billed per token against your credit balance.
- Uncached prompt tokens bill at the input rate.
- Cached prompt tokens bill at the cache-read rate when the model has one.
- Generated tokens, including reasoning tokens, bill at the output rate.
See Usage and Billing for how token counts are reported and charged.
Model retirement
When a model is scheduled for retirement, we announce the date and remove it from the catalog after the retirement date. Requests to a retired model fail with 404 and code model_not_found.
Model IDs
Use the exact id returned by the catalog. An unknown model returns 404 with code model_not_found.