How does Sol's coding performance compare to Claude Mythos?

On Terminal-Bench 2.1, Sol scores 88.8 percent versus Claude Mythos 5's 88 percent, while Sol Ultra reaches 91.9 percent. Sol also shows gains in biology benchmarks, scoring 30 percent on GeneBench v1 versus GPT-5.5's 22 percent best case while using fewer tokens.

When and where will Sol be available?

Sol is currently in limited preview for select partners through the OpenAI API and Codex at the direction of the US government. In July, Sol is set to go live on Cerebras at up to 750 tokens per second.

What are the pricing tiers?

Per million tokens, Sol costs $5 input and $30 output; Terra costs $2.50 and $15; Luna costs $1 and $6. Prompt cache writes cost 1.25x the regular input price, and cache reads receive a 90 percent discount.

Large Language ModelsAI Business & IndustryTHE DECODERJun 26, 2026Jun 27, 2026, 04:00 JST2 min read

OpenAI launches GPT-5.6 Sol, rivals Claude in coding under US government access limits

Send on LINE

Key takeaway

OpenAI has released GPT-5.6 Sol, a new model that edges ahead of Anthropic's Claude Mythos in coding benchmarks while matching it on cybersecurity tasks with greater token efficiency.
The model is available only to select API partners under US government restrictions that OpenAI says should not become long-term policy.
Pricing tiers range from $1–$30 per million tokens depending on the model variant and token type, with a broader launch planned for Cerebras in July.

3 Key Points

What happened
OpenAI unveiled GPT-5.6 Sol, a new flagship model competing with Anthropic's Claude Mythos. Access is limited to select partners through the API and Codex at the US government's direction. The model comes in three tiers—Sol (flagship), Terra (matching GPT-5.5 performance at half the cost), and Luna (budget option)—plus "max" and "ultra" reasoning modes.
Why it matters
On Terminal-Bench 2.1, Sol scores 88.8 percent in agentic coding (AI agents that independently handle complex tasks), edging Claude Mythos 5's 88 percent. Sol Ultra reaches 91.9 percent. On cybersecurity benchmarks, Sol matches Claude's performance while using roughly a third of the output tokens, suggesting lower effective costs per task despite broader inflation in AI pricing. OpenAI has publicly objected to government-mandated access restrictions, calling them unsustainable.
What to watch
Sol pricing is $5 per million input tokens and $30 per million output tokens; Terra costs $2.50 and $15; Luna costs $1 and $6. In July, Sol launches on Cerebras at up to 750 tokens per second. OpenAI has also revamped prompt caching with a guaranteed 30-minute minimum lifetime and 90 percent discounts on cache reads.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

FAQ

How does Sol's coding performance compare to Claude Mythos?: On Terminal-Bench 2.1, Sol scores 88.8 percent versus Claude Mythos 5's 88 percent, while Sol Ultra reaches 91.9 percent. Sol also shows gains in biology benchmarks, scoring 30 percent on GeneBench v1 versus GPT-5.5's 22 percent best case while using fewer tokens.
When and where will Sol be available?: Sol is currently in limited preview for select partners through the OpenAI API and Codex at the direction of the US government. In July, Sol is set to go live on Cerebras at up to 750 tokens per second.
What are the pricing tiers?: Per million tokens, Sol costs $5 input and $30 output; Terra costs $2.50 and $15; Luna costs $1 and $6. Prompt cache writes cost 1.25x the regular input price, and cache reads receive a 90 percent discount.

THE DECODERRead Original Article