GPT-6 Sol is one of the GPT-6 models OpenAI released in September 2026 alongside GPT-6 Luna, at roughly half the price of the equivalent GPT-5.6 models. Recent coverage has centered on comparisons and competition: Vals AI reported DeepSeek V4.1 Flash beat it on a multi-step cybersecurity task, while Laminar's flow-1 and Artemis Security's Orion-1 use it as a benchmark reference. Other stories cover agent-team costs and Codex plan limits.
FireFlash releases recipe for 300B agentic RL model trained in 5 daysResearchers introduced FireFlash, a 300B-scale model for long-horizon agentic tasks, scoring 46.2% on Terminal-Bench 4.0 — on par with Fable 5 and GPT-6 Sol while costing over 30x less per task. It was trained in under 5 days on just 64 NVIDIA B300 GPUs, and the team says it will release the full recipe and open weights.
Vals AI: DeepSeek V4.1 Flash beats GPT 6 Sol on cybersecurity taskVals AI found that DeepSeek V4.1 Flash outperformed GPT 6 Sol on a multi-step offensive cybersecurity task. Sol was tested via the Daybreak trusted access program.
Laminar launches flow-1 model to catch silent agent failuresLaminar, a YC S24 startup, released flow-1, an RL-trained model that finds implicit errors in AI agent traces. The company says it matches GPT-6-sol on detection quality on its own benchmark at 23x lower cost, making it affordable to analyze every production trace without sampling. flow-1 runs in a harness called Signals, treating an agent trace like a code repository and inspecting each span.
Artemis Security launches Orion-1, an AI model purpose-built for cyber defenseArtemis Security released Orion-1, a model purpose-built for cyber defense. It beats Claude Opus 5.5 and GPT-6 Sol on the company's own benchmark. Orion-1 aims to fill the gap as general-purpose frontier models lock down their cyber capabilities, blocking defenders.
Research: AI agent teams cost up to 5.1x more for tiny quality gainsVals AI research finds AI agent teams barely outperform solo agents while costing up to 5.1x more; only one of four tests with GPT-6 Sol and Claude Opus 5.5 showed measurable gains. Anthropic data also shows quality plateauing beyond ten agents as token costs keep rising.
Microsoft launches Microsoft-Decision-1 decision modelMicrosoft introduced Microsoft-Decision-1, a decision-scoring model for routing, classification, prioritization and verification, available in Microsoft Foundry and soon on OpenRouter. Microsoft says it topped a 36-benchmark comparison and was fastest, 4.5x quicker than runner-up Quyet-1.0-Large and 35x than GPT-6 Sol; it was post-trained from Qwen3.5-9B.
A Series of Unfortunate Events for OpenAI UsersAn HN post recounts a rough 30 days for Codex users: after rival Opus 5.5 launched, OpenAI's GPT-6 Sol underperformed, and before DevDay OpenAI cut its $200 plan usage from 20x to 10x, causing user backlash.
OpenAI and Anthropic Cut AI Model Prices, Escalating Price-Performance RaceOpenAI released GPT-6 Sol and Luna with 50% lower costs than predecessors. Anthropic launched Claude Opus 5.5, priced 20% below Opus 5 and performing near top-tier models. Analysts say the race is shifting to efficiency.
Claude Opus 5.5 and GPT-6 Launch Spark Price WarAnthropic released Claude Opus 5.5, and about an hour later OpenAI launched GPT-6 Sol and GPT-6 Luna at roughly half the price of their GPT-5.6 equivalents. GPT-6 Luna costs just $0.10/M input and $0.50/M output, among OpenAI's cheapest models, intensifying the price war.
The A.I. Daily