Models
Model IDs, endpoints, and rates for GPT-6 Astra, Claude, Grok, Kimi, GLM, images, and voice.
Draftworks serves models from five providers through one API key and one prepaid balance. GPT-6 Astra runs through Azure. Claude Fable 5.1, Fable 5, Opus 5, and Sonnet 5 run through Amazon Bedrock.
Catalog
| Model ID | Provider | Context | Endpoints | Tier |
|---|---|---|---|---|
gpt-6-astra | OpenAI | 1,050K | chat.completions, responses | All |
claude-fable-5-1 | Anthropic | 1,000K | chat.completions | All |
claude-opus-5 | Anthropic | 1,000K | chat.completions | All |
claude-sonnet-5 | Anthropic | 1,000K | chat.completions | All |
claude-fable-5 | Anthropic | 1,000K | chat.completions | All |
grok-4.6 | xAI | 500K | chat.completions | All |
kimi-k3 | Moonshot AI | 1,040K | chat.completions | All |
glm-5.3 | Z.ai | 1,000K | chat.completions | All |
gpt-5.6-sol | OpenAI | 1,047.576K | chat.completions, responses | All |
gpt-5.6-terra | OpenAI | 1,047.576K | chat.completions, responses | All |
gpt-5.6-luna | OpenAI | 1,047.576K | chat.completions, responses | All |
gpt-5.4-pro | OpenAI | 1,047.576K | responses | All |
grok-4.3 | xAI | 200K | chat.completions | All |
kimi-k2.7-code | Moonshot AI | 262.144K | chat.completions | All |
kimi-k2.7-code-fast | Moonshot AI | 262.144K | chat.completions | All |
gpt-image-2 | OpenAI | N/A | images.generations | Scale |
gpt-realtime-2.1 | OpenAI | N/A | realtime | Scale |
Use GET /v1/models for the current catalog or GET /v1/models/{id} for one model. A model requested on an unsupported endpoint returns 404 model_not_supported. The API never silently replaces a requested model with a different one.
Text model rates
Prices per million tokens, 40% below provider list on both Base and Scale. These are standard-context rates.
| Model | Input | Cached input | Output |
|---|---|---|---|
gpt-6-astra | $6 | $0.6 | $30 |
claude-fable-5-1 | $6 | $0.15 | $30 |
claude-opus-5 | $3 | $0.3 | $15 |
claude-sonnet-5 | $1.2 | $0.12 | $6 |
claude-fable-5 | $6 | $0.6 | $30 |
grok-4.6 | $1.2 | $0.3 | $3.6 |
kimi-k3 | $1.8 | $0.18 | $9 |
glm-5.3 | $0.84 | $0.156 | $2.64 |
gpt-5.6-sol | $2.4 | $0.24 | $12 |
gpt-5.6-terra | $1.2 | $0.12 | $7.2 |
gpt-5.6-luna | $0.12 | $0.012 | $0.72 |
gpt-5.4-pro | $18 | $18 | $108 |
grok-4.3 | $0.75 | $0.12 | $1.5 |
kimi-k2.7-code | $0.57 | $0.114 | $2.4 |
kimi-k2.7-code-fast | $0.63 | $0.108 | $2.64 |
See the pricing page for provider list prices, Scale rates, image and audio rates, and cache-write prices.
Long contexts and caching
- Astra and GPT-5.6 requests above 272,000 input tokens use 2x input and cached-input rates, and 1.5x output rates for the whole request.
- Grok 4.6 requests at or above 200,000 input tokens use 2x input, cached-input, and output rates.
- Cache creation is billed separately where the provider reports it. Claude cache writes cost 1.25x input for a five-minute entry, or 2x for a one-hour entry. These tokens are included in the total input count and are charged once.
- The same tier discount applies to these rates. Provider prices and availability can change.
Choosing an endpoint
Every text model except GPT-5.4 Pro supports Chat Completions. Astra and the GPT-5.6 family also support Responses. GPT-5.4 Pro requires Responses. Claude supports the documented Bedrock adapter options; it does not expose the native Anthropic Messages endpoint.
Images and realtime voice require the Scale tier, which unlocks at $5,000 in lifetime purchases. Existing model IDs remain available.
Sources
Prices and capabilities were checked on September 5, 2026 against OpenAI, Anthropic, xAI, Fireworks Kimi K3, and Z.ai. Serving deployments are verified in Azure and AWS.