GPT-5.6 Sol vs Fable 5 vs Mythos 5

Quick summary

• On OpenAI's Terminal-Bench chart, Sol Ultra leads clearly; standard Sol and Mythos 5 are essentially tied; Fable 5 sits a notch below.
• All four model numbers are OpenAI-reported — independent leaderboards don't yet list them.
• The timing matters: GPT-5.6 went public July 9 at half Fable 5's token price — three days before Fable 5's plan window was due to close (since extended to July 19).

The chart everyone's sharing

OpenAI's announcement leads with Terminal-Bench 2.1, an agentic-coding benchmark. Here are the scores from that chart:

Read it honestly: Sol Ultra's lead is real on this chart, but ultra is a multi-agent mode that spends far more compute. Standard Sol (88.8) and Mythos 5 (88.0) are within a point. Fable 5 at 84.3 trails by a few points — not a generation.

Why to treat the numbers carefully

Every OpenAI and Anthropic score above comes from OpenAI's own launch chart, not an independent test. The public Terminal-Bench leaderboards don't yet list any of these models, and their current top independently-measured score sits well below Sol's claimed 91.9. Only Gemini 3.1 Pro's 70.7 is independently verified — and it isn't part of OpenAI's comparison set.

There's also no published GPT-5.6 score on cybersecurity's ExploitBench — OpenAI only claims Sol is competitive with Mythos while using about a third of the tokens. Treat all of this as vendor framing until independent results land.

The access flip: Fable 5 is back, GPT-5.6 isn't

The benchmark is a paper race, but access isn't. Fable 5 is back worldwide as of July 1 — you can run it inside Claude, Claude Code, and Cowork (in-plan on Max and Team Premium, on metered usage credits for Pro). GPT-5.6's ~20-partner gate lifted on July 9 — the chart question is now testable on your own work. With Sol at a confirmed $5/$30 versus Fable 5's $10/$50, the cost side of the comparison is not close. Run both before Fable 5's plan window closes July 19.

• GPT-5.6: who can use it and when Access, tiers, pricing, and the controversy
• What Fable 5 costs now the credits math and decision rules

From the Field Guide

Stop guessing which model to use

Weekly verified workflows for the credits era: which model in which seat, what it costs per task, and copy-ready prompts — plus an 8-runbook Starter Pack you can download once and keep. Every claim checked against published pricing, sources shown.