OpenAI just dropped its GPT-5.6 model family, and the flagship Sol variant is already climbing the leaderboards. According to independent evaluations from Artificial Analysis, Sol ranks second in the Intelligence Index, trailing only Anthropic’s Claude Fable 5, while simultaneously claiming the top spot in the Coding Agent Index.

The kicker: it does this at roughly a third of what Claude Fable 5 charges.

The numbers behind Sol’s rise

Sol scored 88.8% on Terminal-Bench 2.1 in standard mode. In its “Ultra” configuration, which deploys multiple sub-agents working in concert, that number jumps to 91.9%.

On the pricing side, Sol comes in at $5 per million input tokens and $30 per million output tokens. Claude Fable 5, its closest competitor in the intelligence rankings, charges $10 input and $50 output per million tokens.

The model also supports inference speeds of up to 750 tokens per second on Cerebras hardware.

Sol doesn’t arrive alone. It’s part of a trio alongside Terra and Luna, though Sol is clearly the headliner. The limited preview launched on June 26, 2026, with general public access expected around July 6-9. OpenAI plans to integrate the model into its Codex platform.

The crypto angle nobody asked for

OpenAI named its three models Sol, Terra, and Luna. SOL is Solana’s native token. Terra and Luna are the names of the stablecoin ecosystem that spectacularly collapsed in May 2022, wiping out roughly $40 billion in value and triggering contagion across the entire digital asset market.

There is no reported connection between OpenAI’s naming choices and any cryptocurrency project. The names appear to follow a celestial theme, not a blockchain one.

Why AI model economics matter for crypto

AI-powered coding agents are increasingly used to build, audit, and deploy smart contracts. When the cost of deploying these tools drops by two-thirds while performance stays flat or improves, adoption accelerates.

The regulatory backdrop adds another layer. OpenAI released GPT-5.6 Sol as a limited preview restricted to trusted partners, partly due to increased US government scrutiny over advanced AI capabilities. Safety evaluations of Sol noted a slight uptick in misalignment and what OpenAI characterized as “cheating” behaviors compared to GPT-5.5, though these incidents were described as rare.

What investors should watch

The July 6-9 public rollout date is the next catalyst to watch. Once Sol becomes widely available, real-world performance data will start flowing from developers outside OpenAI’s trusted partner circle.