OpenAI's new GPT-5.6 Sol model performs nearly as well as Anthropic's Fable 5 in benchmark tests, according to independent AI evaluation platform Artificial Analysis. The model scores 59 points on the Intelligence Index, one point behind Fable 5's 60. In the Coding Agent Index, Sol (max) leads with 80 points in OpenAI's Codex environment, outperforming other models. The model's cost is significantly lower, at $1.04 per task, about one-third of Fable 5's $2.75. Smaller variants, Terra and Luna, cost $0.55 and $0.21 per task, respectively. OpenAI CEO Sam Altman noted that Sol 'defines a new Pareto frontier of Intelligence vs Output Tokens per Task.'
GPT-5.6 Sol introduces a cache-write fee, further reducing costs. Token prices per million are $5/$30 for input/output, $2.50/$15 for Terra, and $1/$6 for Luna. Cache reads receive a 90 percent discount. Sol also burns up to 54 percent fewer output tokens than models with similar performance, particularly in agentic coding tasks. These cost efficiencies position Sol as a competitive option in the AI market. The model's pricing strategy is part of a broader industry trend toward lower costs, with competitors like Chinese open models, Meta with Muse 1.1, and xAI with Grok 4.5 pushing established labs to adjust their rates.
Artificial Analysis provided the benchmark data, highlighting Sol's performance across multiple metrics. The full results are available on the platform's website. OpenAI's pricing strategy is seen as a direct challenge to Anthropic, which must now respond to the competitive pressure. This pricing race could have implications for the entire AI industry. The company's approach reflects ongoing efforts to balance performance and affordability in a rapidly evolving market.
Source: thedecoder
