OpenAI's new GPT-5.6 family of artificial intelligence models, released on July 9, 2026, has shown strong performance against Anthropic's Fable Class, with its flagship GPT-5.6 Sol model edging out Anthropic's Claude Mythos 5 on a critical coding benchmark. The new OpenAI models also present a significantly more cost-effective option for developers and businesses.[en+3]
New Models Face Government Scrutiny
Both OpenAI and Anthropic recently launched their latest powerful AI models after facing review from the US government. OpenAI released its GPT-5.6 models following a preview period for trusted partners.This staggered release came after the Trump administration requested AI developers to share access to new frontier models for government review before public release.Anthropic's Fable 5 and Mythos 5 models also experienced temporary export bans and restrictions due to similar government concerns over cybersecurity risks.[theguardian+10]
OpenAI's Triple Threat: Sol, Terra, and Luna
The GPT-5.6 family includes three distinct models: Sol, Terra, and Luna.Sol stands as the most capable and flagship model, designed for complex reasoning, coding, scientific research, and cybersecurity tasks.Terra offers a balanced, intermediate option, while Luna is the fastest and most budget-friendly choice.OpenAI describes Sol as its "workhorse" and "best coding model yet."[en+9]
On the TerminalBench 2.1 benchmark, GPT-5.6 Sol achieved an 88.8% score, surpassing Claude Mythos 5, which scored 88.0%.An even more powerful variant, Sol Ultra, posted an impressive 91.9% on the same benchmark.The more economical GPT-5.6 Luna also outperformed Anthropic's Claude Opus 4.8 on TerminalBench 2.1, scoring 82.5% compared to 78.9%.[en+2]
Anthropic's Fable 5 Leads Overall Leaderboard
Anthropic countered with its Mythos-class models, Fable 5 and Mythos 5, announced on June 9, 2026.Fable 5 is Anthropic's most capable widely released model, excelling in software engineering, knowledge work, vision, and scientific research.It is particularly optimized for reliable tool use and following instructions across long, multi-step tasks.[hidekazu-konishi+7]
As of July 10, 2026, Claude Fable 5 leads the overall BenchLM leaderboard with a score of 91, followed by Claude Mythos 5 at 89.This indicates Fable 5's strong general performance across various workloads. Claude Mythos 5 is the same underlying model as Fable 5 but with certain safety safeguards lifted, primarily for select government and cybersecurity partners through Project Glasswing.[benchlm+3]
Cost and Accessibility Differences
OpenAI's GPT-5.6 models generally offer lower pricing per token compared to Anthropic's Fable 5.GPT-5.6 Sol is priced at $5 per million input tokens and $30 per million output tokens.Its mid-tier model, Terra, costs $2.50 per million input tokens and $15 per million output tokens, while Luna is the most affordable at $1 per million input and $6 per million output.[mashable+10]
In contrast, Anthropic's Claude Fable 5 and Mythos 5 are both priced at $10 per million input tokens and $50 per million output tokens.This makes OpenAI's offerings, particularly Terra and Luna, significantly more cost-effective for many applications, especially those with high output volumes.[engadget+1]
GPT-5.6 Sol is currently available to users with paid ChatGPT plans, including Plus, Business, and Enterprise.All three GPT-5.6 variants are also accessible via the OpenAI API, Codex, and GitHub Copilot.Claude Fable 5 is generally available through the Claude API and platforms like Amazon Bedrock and Google Cloud.[notebookcheck+4]
Safety and Future Outlook
Both companies emphasize safety in their new releases. OpenAI's GPT-5.6 family is classified with "High capability" in cybersecurity and biological/chemical risk domains under OpenAI's Preparedness Framework.However, OpenAI notes that these models cannot autonomously launch end-to-end cyberattacks or create zero-day exploits without human involvement.Fable 5 also incorporates robust safeguards, automatically routing high-risk cybersecurity and biology queries to the less capable Opus 4.8 model.[en+6]
The ongoing competition between OpenAI and Anthropic continues to drive rapid advancements in AI capabilities. While Anthropic's Fable 5 holds a lead on some broad benchmarks, OpenAI's GPT-5.6 Sol demonstrates superior performance in specific areas like coding and offers a more budget-friendly option across its model family. This sets the stage for intensified competition in enterprise AI solutions as businesses seek both raw intelligence and cost efficiency.
