GPT-6 Astra API Pricing Set at $10 and $50 Per Million Tokens

Standard tier matches the frontier price ceiling

When GPT-6 Astra becomes available through the OpenAI API under the model identifier gpt-6-astra, standard pricing will be $10 per million input tokens and $50 per million output tokens. Cached input reads cost $1 per million tokens, and cache writes are billed at $12.50 per million. Fast mode, which OpenAI says delivers up to twice the speed of standard processing, doubles those rates to $20 and $100.

The list price is 2.5 times GPT-5.6 Sol's current promotional pricing of $4 and $20 per million tokens, but it aligns exactly with Anthropic's Claude Fable 5.1, which also charges $10 and $50 at standard rates. Batch and Flex tiers run at 50% of standard pricing, offering $5 and $25 per million for workloads that tolerate delayed processing.

Long prompts and regional surcharges add cost

Developers face additional pricing complexity on large-context workloads. Requests exceeding 272,000 input tokens are billed at double the input and cache rates and 1.5 times the output rate for the entire request, not just the portion above the threshold. With Astra's 1.05-million-token context window, teams running full-codebase or long-document tasks need to model costs carefully.

Regional processing endpoints carry a 10% uplift, and Astra is unavailable on the free API tier, meaning Tier 1 is the minimum entry point. OpenAI also offers Astra through Amazon Bedrock, though partner pricing may differ from first-party API rates.

Whether higher per-token rates mean higher bills

OpenAI argues that Astra's higher per-token price does not necessarily translate to higher total cost. The company says the model completes several benchmark tasks using fewer output tokens and with fewer retries than earlier systems. Partner pilots cited roughly 40% improvement on financial-statement review and 50% less manual fix work on game prototypes.

Independent analysts caution that launch data is too sparse to confirm those savings offset the price premium at scale. For agentic workloads that accumulate large repeated context, Anthropic's Fable 5.1 offers a structural advantage: cache reads at $0.25 per million tokens versus Astra's $1, a difference that can materially affect bills on long-running coding agents.

Subscription access included in existing allowances

For ChatGPT subscribers, Astra usage is included within existing subscription allowances, with the option to purchase additional credits. Pro, Business, and Enterprise plans also receive access to GPT-6 Astra Pro, a variant with extended capabilities. The pricing announcement therefore affects API and cloud customers most directly, while consumer and business chat users see the model as part of their existing plan economics.