OpenAI's GPT-6.1 Sol Arrives: Astra-Level Capabilities at a Fraction of the Price

OpenAI has unveiled GPT-6.1 Sol, a significant upgrade to its GPT-6 Sol model. This new addition positions itself as a robust mid-tier solution within the GPT-6 ecosystem, offering capabilities that rival the top-tier Astra model. Developers and businesses can now access advanced AI functionalities for agentic coding, computer interaction, and various professional tasks, all while benefiting from a dramatically lower token price.

Pricing and Availability

The GPT-6.1 Sol model is designed to deliver exceptional value. Standard input and output rates for GPT-6.1 Sol are set at one-fifth the cost of the premium GPT-6 Astra ($2 input, $10 output per million tokens). Cached input pricing notably improves to $0.10 per million tokens, a 50% reduction from its predecessor. This model is immediately available for integration via the OpenAI API (gpt-6.1-sol), ChatGPT Work, and Codex.

The GPT-6 Family Overview

OpenAI’s GPT-6 model range now consists of three distinct tiers:

• GPT-6 Astra: The flagship model ($10 input, $50 output, $1 cached input per million tokens).
• GPT-6.1 Sol: The new mid-tier option ($2 input, $10 output, $0.10 cached input per million tokens).
• GPT-6 Luna: The most economical tier ($0.10 input, $0.50 output, $0.01 cached input per million tokens).

The reduced cost for cached inputs is particularly advantageous for AI agents, which frequently resend identical system prompts. Cached reads now cost only 5% of the uncached input rate, down from 10% on GPT-6 Sol.

Performance Highlights

OpenAI's internal benchmarks indicate that GPT-6.1 Sol performs remarkably well across several key domains, underscoring its efficiency and power.

• Coding Prowess: On DeepSWE v1.1, GPT-6.1 Sol achieved performance levels on par with GPT-6 Astra, at approximately one-fifth of the operational cost. It also surpassed GPT-6 Sol’s prior top score by 6.4 percentage points.
• Professional Task Handling: In tests involving complex professional PDFs (GDP.pdf), Sol outperformed Claude Opus 5.5, often at less than half the task cost. On AutomationBench 1.0.6, Sol registered a score 2.2 points higher than Opus 5.5, at roughly one-third of the expense.
• Computer Interaction: On the OSWorld 2.0 offline dataset, Sol demonstrated a 7-point gain over GPT-6 Sol. Its results placed it within 2.1 points of Astra, while costing roughly one-seventh per task.
• Scientific Applications: On Terminal-Bench Science 0.1, Sol significantly improved upon GPT-6 Sol’s score. While Astra maintains its lead for the most demanding research tasks, Sol offers a compelling alternative at an average task cost of $5.47, compared to over $23 for Opus 5.5 and Astra.
• Enhanced Factuality: The model also shows improved factual accuracy. In challenging conversational scenarios, the rate of factual errors decreased by approximately 32%, falling from 11.4% to 7.7%.

Key Developer Specifications

Developers leveraging GPT-6.1 Sol through the API will find a robust set of features:

• A substantial context window of 1,050,000 tokens and 128,000 max output tokens.
• A knowledge cutoff date of April 30, 2026.
• Support for text and image inputs, with text-only outputs.
• Flexible reasoning.effort settings (low to max, default medium).
• Prompts exceeding 272,000 input tokens incur a surcharge (2x input/cache, 1.5x output).
• Batch and Flex processing options offer a 50% cost reduction.
• US and EU data residency are supported; Fast mode is incompatible with EU residency.
• Fine-tuning capabilities are not supported.

OpenAI has also indicated plans for a GPT-6.1 Sol Ultrafast option within Codex soon, promising generation speeds up to eight times faster than standard rates.

Competitive Landscape

GPT-6.1 Sol enters a competitive market, positioning itself against models like Claude Opus 5.5, Claude Sonnet 5.5, and Gemini 3.1 Pro Preview. While Claude Sonnet 5.5 matches Sol's primary input/output pricing, Sol offers a significantly cheaper cached input. OpenAI's performance comparisons primarily feature Claude Opus 5.5, highlighting Sol's strong value proposition.