technology•Wed, September 30, 2026•5 min read

OpenAI Just Changed the Economics of AI Agents with GPT-6.1 Sol

OpenAI has unveiled GPT-6.1 Sol, a powerful new model engineered to deliver performance nearing its flagship Astra tier at just 20% of the cost. Designed specifically for professional coding and complex agentic tasks, this release signals a major shift in how developers should balance intelligence, latency, and budget.

A New Benchmark for Cost-Efficiency

In a move that caught many developers by surprise, OpenAI has replaced the week-old GPT-6 Sol with GPT-6.1 Sol. Announced during the annual OpenAI DevDay 2026, the new model is designed to sit comfortably between the flagship GPT-6 Astra and the high-speed GPT-6 Luna. For enterprise leaders and developers building agentic workflows, the implications are significant: you can now access near-Astra level intelligence for one-fifth of the standard API cost.

GPT-6.1 Sol bridges the gap between flagship performance and operational budget.

Why GPT-6.1 Sol Matters for AI Agents

The shift toward 'agentic' computing—where AI models perform multi-step tasks like coding in a terminal or navigating complex professional documents—requires models that are both smart and cost-effective. GPT-6.1 Sol excels here by offering a superior price-to-performance ratio for demanding tasks.

Price: API rates are set at $2 per million tokens in and $10 per million tokens out, significantly undercutting Astra's $10 and $50 rates.
Caching Efficiency: Input caching costs have been slashed to $0.10 per million tokens, a 95% reduction from standard rates, which is a game-changer for long-context agent loops.
Performance: Benchmarks show GPT-6.1 Sol outperforming the previous Sol iteration and matching Astra across several high-stakes professional evaluations.
Tiered Flexibility: The release also introduces an 'Ultrafast' tier for applications where low latency is the priority over cost.

OpenAI is increasingly asking developers to optimize not simply for which model is smartest, but across three separate variables: intelligence, cost and latency.
— VentureBeat

The Future of Professional AI Workflows

For developers, this isn't just about a cheaper model; it's about the ability to scale. By significantly lowering the barrier to entry for high-performance agentic tasks—such as software engineering in large codebases—OpenAI is enabling more complex, autonomous workflows to become economically viable. With benchmarks showing GPT-6.1 Sol clearing 75% on DeepSWE, the model is clearly aimed at professionals who need reliable, high-level reasoning without the premium price tag of a flagship model.

Key Takeaways

GPT-6.1 Sol offers near-Astra performance at one-fifth of the standard API costs.
The model is optimized specifically for coding, computer use, and professional document workflows.
Cache costs for long-running agent loops are reduced by 95% to $0.10 per million tokens.
OpenAI introduced a new 'Ultrafast' tier for latency-sensitive applications.
The model replaces the original GPT-6 Sol, effectively setting a new standard for mid-range, high-intelligence models.

FAQ

How much does GPT-6.1 Sol cost?

It is priced at $2 per million input tokens and $10 per million output tokens, which is five times cheaper than the flagship GPT-6 Astra model.

Is GPT-6.1 Sol better than GPT-6 Astra?

While Astra remains the flagship model, GPT-6.1 Sol offers near-Astra performance for a fraction of the cost, making it highly competitive for most professional and agentic tasks.

What is the new 'Ultrafast' tier?

The Ultrafast tier is designed for applications where low latency is critical, allowing for speeds of up to 300 tokens per second for a premium cost.

Why did OpenAI release GPT-6.1 Sol so quickly after the original Sol?

The 6.1 version provides significant performance improvements and better cost-efficiency, effectively replacing the original model to better meet the needs of agentic and coding-heavy workloads.

Does GPT-6.1 Sol support cached tokens?

Yes, it supports input caching at a significantly reduced rate of $0.10 per million tokens, which helps lower costs for recurring agent tasks.