{DevDay 2026} GPT-6.1 Sol moves closer to Astra without adopting its pricing

GPT-6.1 Sol improves on GPT-6 Sol across agentic coding, Computer Use, professional work, and scientific research. OpenAI keeps pricing at $2 per million input tokens and $10 per million output tokens, one fifth of GPT-6 Astra's standard rates.

Closing the gap with Astra

Between its everyday and frontier models, OpenAI is narrowing the performance gap. GPT-6.1 Sol retains Sol-level pricing while moving closer to GPT-6 Astra across several evaluations focused on agentic and professional work.

Standard API pricing remains $2 per million input tokens and $10 per million output tokens. Cached input drops to $0.10 per million tokens, 95% below the standard input rate.

GPT-6 Astra, by comparison, costs $10 per million input tokens and $50 per million output tokens. GPT-6.1 Sol therefore runs at one fifth of those two standard rates, although OpenAI continues to position Astra as its strongest overall model. Agentic coding reaches Astra on DeepSWE

On DeepSWE v1.1, an evaluation of long-horizon software engineering tasks in real codebases, GPT-6.1 Sol matches GPT-6 Astra in OpenAI's testing at roughly one fifth of the cost per task.

It also exceeds GPT-6 Sol's best score by 6.4 percentage points while using a lower reasoning effort and costing less.

The improvement extends to professional workflows. On AutomationBench 1.0.6, where agents use 47 tools to complete end-to-end workflows across sales, marketing, operations, support, finance, and HR, GPT-6.1 Sol outperforms GPT-6 Sol by 4.8 percentage points at the same reasoning setting.

OpenAI also reports a 2.2-point advantage over Opus 5.5 with fallbacks at Medium reasoning effort, at roughly one third of the cost per task. Complex PDFs enter the equation

GDP.pdf evaluates how models answer professional questions using complex PDFs containing tables, charts, diagrams, and fine print across ten fields including finance, healthcare, and law.

In OpenAI's published evaluations, GPT-6.1 Sol outperforms Opus 5.5 with fallbacks at less than half the cost per task across every tested reasoning setting.

Its results also approach GPT-6 Astra's reference performance at roughly one fifth of the cost per task. Within 2.1 points of Astra on Computer Use

Interaction with desktop applications also improves. On the offline set of OSWorld 2.0, GPT-6.1 Sol gains seven percentage points over GPT-6 Sol at maximum reasoning effort while costing less than half as much per task.

At the same effort level, it sits 2.1 percentage points behind GPT-6 Astra while costing roughly one seventh as much per task, according to OpenAI.

The evaluation covers long-horizon Computer Use workflows spanning both everyday and professional tasks. Astra still leads on scientific research

The gap remains more visible on the most demanding scientific work. Terminal-Bench Science 0.1 evaluates agents on workflows involving data analysis, simulation, model fitting, and theorem proving through code and command-line tools.

GPT-6.1 Sol more than doubles GPT-6 Sol's score at maximum reasoning effort. OpenAI reports an average cost of $5.47 per task at that setting, compared with $23.21 for Opus 5.5 and $23.80 for GPT-6 Astra.

Astra still records the highest score in the comparison at 68.1% and remains OpenAI's recommended model for its most demanding scientific research tasks. A 4.1% error rate on deliberately difficult prompts

OpenAI also evaluates factual accuracy using de-identified ChatGPT conversations in which users had previously reported a factual error.

At very high reasoning effort, 4.1% of GPT-6.1 Sol responses contain at least one factual error, compared with 4.5% for GPT-6 Sol and 4% for GPT-6 Astra.

OpenAI notes that the dataset was deliberately constructed from error-prone conversations and is not representative of the frequency of factual errors in ordinary usage. Available in Work, Codex, and the API

GPT-6.1 Sol is available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. It is not yet available in Chat.

Developers can access the model through the API under the `gpt-6.1-sol` model ID.

OpenAI is also introducing GPT-6 Astra Ultrafast and GPT-6.1 Sol Ultrafast, which it says can run up to eight times faster than GPT-6 Astra. The variants are available through the API, ChatGPT Work, and Codex under a new $500-per-month Pro tier.