OpenAI announces GPT-5.6 with three models (Sol, Terra, Luna) and ChatGPT Work agent tool
OpenAI released GPT-5.6 in three model tiers—Sol (flagship reasoning), Terra (mainstream), and Luna (instant)—positioning them against Anthropic's Claude models. The company claims GPT-5.6 Sol scores 53.6 on Agents' Last Exam, 13.1 points above Claude Fable 5, while completing tasks 61% faster. ChatGPT Work, a desktop productivity agent similar to Claude Cowork, launches simultaneously for Pro, Enterprise, and Edu users.
GPT-5.6 Sol — Quick Specs
OpenAI announces GPT-5.6 with three models (Sol, Terra, Luna) and ChatGPT Work agent tool
OpenAI released GPT-5.6 today in three model tiers, directly targeting Anthropic's Claude lineup with claims of superior performance at lower cost. The company also launched ChatGPT Work, a desktop productivity agent that competes with Claude Cowork.
Three model tiers with celestial names
GPT-5.6 launches in three configurations:
- Sol: Flagship reasoning model, equivalent to GPT-5.5's thinking mode, aimed at Claude Fable 5
- Terra: Default mainstream model, equivalent to GPT-5.5 standard
- Luna: Faster, less capable model, equivalent to GPT-5.5's instant mode
The release follows a government-mandated delay last month related to safety reviews, coming after the "Anthropic/White House kerfuffle" over Claude Fable 5 and Mythos 5.
Benchmark claims target Anthropic directly
OpenAI's announcement explicitly compares GPT-5.6 Sol to Claude Fable 5 across multiple benchmarks:
- Agents' Last Exam (tests long-running workflows across 55 professional fields): Sol scored 53.6, claimed to be 13.1 points above Fable 5
- Artificial Analysis Intelligence Index: Sol came within one point of Fable 5 while completing tasks in 61% less time, according to OpenAI
- Artificial Analysis Coding Agent Index: Sol reached 80, claimed to be 2.8 points above Fable 5
OpenAI also claims Sol beat Fable 5 by 11.4 points at medium reasoning while costing roughly one-quarter the price. Specific pricing figures were not disclosed.
Qodo, an AI code review tool, reported that GPT-5.6 "beat GPT-5.5 on F1 while using roughly 3x fewer tokens per PR and delivering about 2x lower median latency."
ChatGPT Work: Desktop productivity agent
ChatGPT Work allows AI access to desktop applications and browsers to perform multi-step tasks. Previously embedded in OpenAI's Codex programming tool, it's now separated as a standalone product.
Capabilities include:
- Creating spreadsheets, presentations, and websites
- Multi-hour complex project execution by breaking into subtasks
- Integration with Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, CRMs, and project trackers
Angela Ferrante, head of enterprise marketing at Zapier, stated the company used ChatGPT Work to review thousands of monthly leads, tracing customer touchpoints and generating executive dashboards that "revealed seven figures in potential sales."
Availability and platform changes
ChatGPT Work launches today on web and mobile for Pro, Enterprise, and Edu plans, with Plus ($20/month) and Business tiers following "over the next few days."
OpenAI is combining its separate ChatGPT and Codex desktop apps into one unified application, similar to Anthropic's single-app approach for Claude chat, Claude Code, and Claude Cowork.
Safety claims
OpenAI states GPT-5.6 includes "our most robust safeguards to date," combining model protections, real-time checks, monitoring, and "access calibrated to trust and risk."
What this means
This release represents OpenAI's most direct competitive response to Anthropic to date, with benchmark comparisons dominating the announcement. The three-tier naming scheme (Sol/Terra/Luna) mirrors industry trends toward clearer product differentiation. ChatGPT Work's separation from coding tools suggests OpenAI is targeting enterprise productivity workflows beyond software development—a market where Anthropic's Cowork has gained traction. The claimed speed and cost advantages, if verified in independent testing, could shift enterprise purchasing decisions, though specific pricing details remain undisclosed.
Related Articles
Chinese Models Kimi K3 and GLM-5.3 Close In on GPT-5.5 and Claude Opus 5, New Analysis Finds
A new industry analysis argues the performance gap between Chinese and Western AI models has narrowed to single-digit differences on broad benchmarks. Moonshot's Kimi K3 and Zhipu's GLM-5.3 now trail OpenAI and Anthropic's top models by only a few points on the Artificial Analysis Intelligence Index, with a clear Western edge remaining only in abstract reasoning, output reliability, and offensive cybersecurity capability.
OpenAI Launches 'Private Safety Processing' to Detect Misuse Without Storing Enterprise Data
OpenAI has built a system called Private Safety Processing that detects misuse patterns across multiple interactions without storing customer inputs or outputs. The company says it only receives narrow safety signals—type and severity of activity—while data stays encrypted on customer infrastructure.
OpenAI Previews 'Private Safety Processing' to Detect Abuse Without Retaining Customer Data
OpenAI is previewing Private Safety Processing to select customers, an automated system that monitors for misuse across multiple sessions without retaining any customer data. The move directly contrasts with Anthropic's July policy allowing 30-day data retention for 'covered models' like Fable.
DeepSeek Releases Experimental V4-Flash-Vision-Exp, Claims Near-Parity With Opus 4.8 on Agent Benchmarks
DeepSeek has released V4-Flash-Vision-Exp, an experimental multimodal extension of V4-Flash that adds image understanding while preserving text reasoning capabilities. The company claims the model approaches or beats Anthropic's Opus 4.8 on its internal multimodal agent benchmarks.
Comments
Loading...