GPT-5.6 Sol nearly matches Fable 5 on aggregated benchmarks at one-third the cost
Quick Answer
OpenAI's GPT-5.6 Sol nearly matches Fable 5 in benchmarks, scoring 59 points versus 60, while costing only $1.04 per task—one-third the price of Fable 5.
Quick Take
OpenAI's GPT-5.6 Sol nearly matches Fable 5 in benchmarks, scoring 59 points versus 60, while costing only $1.04 per task—one-third the price of Fable 5. This pricing strategy pressures competitors like Anthropic and could reshape the AI market landscape.
Key Points
- GPT-5.6 Sol scores 59 on the Intelligence Index, just behind Fable 5's 60.
- Sol costs $1.04 per task, significantly cheaper than Fable 5's $2.75.
- In the Coding Agent Index, Sol ranks first with 80 points in Codex.
- Smaller variants Terra and Luna are 50% and 80% cheaper than Sol, respectively.
- Pricing pressure from competitors could impact the entire AI industry.
📖 Reader Mode
~2 min readOpenAI's new flagship model, GPT-5.6 Sol, performs almost as well as Fable 5 in benchmarks but costs only one-third as much.
According to the independent AI evaluation platform Artificial Analysis, GPT-5.6 Sol (max) scores 59 points on the Intelligence Index, just one point behind Claude Fable 5 (max). In Artificial Analysis's new Coding Agent Index, Sol (max) takes the top spot with 80 points when running in OpenAI's Codex environment, beating every other model. In the AA-Briefcase benchmark for realistic office tasks, Sol earns the highest "Presentation Elo" of all models but still trails Fable 5 in the overall ranking.
OpenAI takes on Anthropic on price
Sol costs $1.04 per task, about one-third of what Fable 5 charges. The smaller variants, Terra and Luna, cost 50 and 80 percent less than Sol, respectively. According to Artificial Analysis, Sol "defines a new Pareto frontier of Intelligence vs Output Tokens per Task." The full benchmark data is available here.
GPT-5.6 also introduces a cache-write fee for the first time, pushing prices down further. Token prices per million are $5/$30 (Sol), $2.50/$15 (Terra), and $1/$6 (Luna) for input/output. Cache reads get a 90 percent discount. Sol also burns fewer output tokens than models with similar performance, up to 54 percent fewer for agentic coding tasks, according to OpenAI CEO Sam Altman.
Pricing pressure is building across the AI industry. Chinese open models, Meta with Muse 1.1, and xAI with Grok 4.5 are all pushing established AI labs to cut rates. OpenAI is still pricier than those competitors but is now squeezing Anthropic from another angle. Anthropic will need to respond. But this race to the bottom on pricing could end up hurting the entire industry.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.
