OpenAI Launches GPT-6 Astra, Challenging the Generative AI Landscape

OpenAI has announced the release of its latest language model, GPT-6 Astra, positioning it as a direct competitor to Anthropic’s Claude Fable. The model is rolling out in phases, starting with select organizations and expanding to all ChatGPT Plus subscribers over the coming days.

Impressive Benchmark Results

Astra has demonstrated exceptional performance across various benchmarks:

• ARC-AGI 3: Achieved a score of 99.9% (using OpenAI’s Provider Adapter harness) - significantly outperforming other models
• ExploitBench: Scored 100%, surpassing GPT-5.6 Sol at 78.5%
• SRE-Bench: Attained 99.2% within four attempts, compared to Sol’s 68.7%

These results indicate a substantial advancement in security and vulnerability analysis capabilities.

Long Context Processing Excellence

Astra excels in handling extended inputs, achieving:

• 100% accuracy on OpenAI’s eight-needle benchmark at 256K–512K tokens
• 96.3% accuracy at 512K–1M tokens

This addresses a common limitation in earlier large language models and opens new applications for processing extensive documents and conversations.

Pricing and Availability

The GPT-6 Astra API will be priced competitively at:

• $10 per million input tokens
• $50 per million output tokens

This aligns with Anthropic’s pricing structure while offering comparable or superior performance across key metrics.

While comprehensive evaluations are ongoing, early data suggests GPT-6 Astra represents a significant leap forward in generative AI capabilities.