OpenAI: GPT-6 Sol (batch)

GPT-6 Sol (batch) is OpenAI's cost-efficient GPT-6 tier aimed at complex coding, tool use, and long-horizon agentic workflows. It inherits alignment and capability improvements from GPT-6 Astra while targeting substantially lower latency and price than the flagship model. The batch API variant exposes the same gpt-6-sol weights with asynchronous processing at 50% of standard input and output token rates, making it suited to large offline evaluation, data generation, and sustained agent runs where throughput matters more than real-time latency.

On OpenAI's launch evaluations, GPT-6 Sol at max or xhigh reasoning reaches strong scores on professional automation (AutomationBench), agentic professional tasks (Agents' Last Exam), and software engineering (DeepSWE v1.1), often matching or beating similarly priced frontier competitors at a fraction of the cost per task. Third-party leaderboards report competitive results on knowledge and reasoning suites (MMLU-Pro, GPQA-Diamond, HLE), LiveBench, SWE-bench Verified and Multilingual, terminal operations (Terminal-Bench 2.1), and creative writing (EQ-Bench Creative Writing v3 Elo).

The model supports text and image inputs with a very large context window, configurable reasoning effort from none through max, and the Responses API tool ecosystem (web search, code interpreter, computer use, MCP, and related agent tools). It is available via the OpenAI API, ChatGPT Work, and Codex for paid tiers; weights are not published for local deployment.