Models
Use a model by its canonical id. AIx handles routing and failover behind the scenes, and billing stays on the model’s published price regardless of the active route.
Coding picks
anthropic/claude-sonnet-4-6— the quality-first all-rounder for agentic coding.moonshotai/Kimi-K2.7-Code— built for tool-using, repository-scale coding.deepseek-ai/DeepSeek-V3.2— strong coding performance at a low unit cost.
Start with Sonnet for quality, Kimi for coding-agent workflows, or DeepSeek for high-volume background tasks. Verify the capability you need in the live catalog before production use.
Comparison
| Model | Type | Context | Price | Best for |
|---|---|---|---|---|
| Claude Opus 4.8anthropic/claude-opus-4-8 | Chat | 200K | $12.5 / 1M tokens | Hardest reasoning & long agentic sessions |
| Claude Sonnet 4.6anthropic/claude-sonnet-4-6 | Chat | 200K | $7.5 / 1M tokens | Balanced coding & agents — the default pick |
| Kimi K2.6moonshotai/Kimi-K2.6 | Chat | 128K | $4.5 / 1M tokens | Agentic workflows & heavy tool use |
| Gemini 2.5 Progoogle/gemini-2.5-pro | Chat | 1M | $4.2969 / 1M tokens | Massive context & multimodal analysis |
| Kimi K2.7 Codemoonshotai/Kimi-K2.7-Code | Chat | 256K | $4 / 1M tokens | Agentic, tool-using software engineering |
| Claude Haiku 4.5anthropic/claude-haiku-4-5 | Chat | 200K | $2.5 / 1M tokens | Fast agent turns and high-volume coding support |
| DeepSeek R1 0528deepseek-ai/DeepSeek-R1-0528 | Chat | 128K | $2.15 / 1M tokens | Open reasoning workloads |
| GLM 4.6zai-org/GLM-4.6 | Chat | 200K | $1.74 / 1M tokens | Coding & agents with strong price/performance |
| Gemini 2.5 Flashgoogle/gemini-2.5-flash | Chat | 1M | $1.0625 / 1M tokens | Fast, cheap, very long context |
| Llama 3.3 70B Instruct Turbometa-llama/Llama-3.3-70B-Instruct-Turbo | Chat | 128K | $1.04 / 1M tokens | Open general-purpose workhorse |
| GPT OSS 120Bopenai/gpt-oss-120b | Chat | 128K | $0.6 / 1M tokens | Open general-purpose reasoning |
| DeepSeek V3.2deepseek-ai/DeepSeek-V3.2 | Chat | 128K | $0.38 / 1M tokens | Strong open-model coding at very low cost |
| DeepSeek V4 Flashdeepseek-ai/DeepSeek-V4-Flash | Chat | 128K | $0.2 / 1M tokens | Fast general and coding workloads |
| FLUX 1.1 Problack-forest-labs/FLUX-1.1-pro | Image | — | $0.04 / image | High-fidelity image generation |
| Kokoro 82Mhexgrad/Kokoro-82M | Speech | — | $0.62 / 1M tokens | Natural, low-cost text-to-speech |
| Whisper Large V3openai/whisper-large-v3 | Transcription | — | $0.0005 / min | Accurate multilingual transcription |
| BGE Large EN v1.5BAAI/bge-large-en-v1.5 | Embeddings | 512 | $0.01 / 1M tokens | General-purpose retrieval embeddings |
Prices update live from GET /api/v1/models; context windows are each
model's published maximum and may vary by active route. See the full catalog →
List models programmatically
The catalog is served live (no SDK required) and updates without a redeploy:
curl https://api.aix.theaimart.co/v1/models
{
"object": "list",
"data": [
{ "id": "deepseek-ai/DeepSeek-V3.2", "object": "model", "owned_by": "theaimart", "type": "chat" },
{ "id": "anthropic/claude-sonnet-4-6", "object": "model", "owned_by": "theaimart", "type": "chat" }
]
}
See the Models API reference for the full response shape, and the public catalog for every available id with live pricing.
How routing works
Each model has an internal set of eligible routes. AIx evaluates them against:
- Capability fit — the route must support the requested input and features.
- Health — degraded or unavailable routes are skipped.
- Efficiency — cost and latency inform selection between healthy routes.
If the selected route errors, times out or its circuit breaker is open, AIx automatically tries the next eligible path. Your integration stays on one stable endpoint. Full detail in Routing & failover.