I'll be using the officially reported Fable 5/Mythos scores from Anthropic.
Yes, I'm aware that Fable and Mythos aren't exactly the same models but Anthropic is treating them as the same so I'll treat them as the same model
I'll be using scores reported by OpenAI on becnhmarks + the score on DeepSWE (if the score exists) + LMArena rankings + other sources and opinions from Theo t3 or Simon Willison.
Tldr: will gather a buncha sources to make my judgement.
This question strictly asks "will it be BETTER", not "will it be on-par"
If openai provides two sets of scores for the gpt5.6 model, one for the censored one and one for the uncensored one, I'll take whichever score is higher
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ25 | |
| 2 | Ṁ23 | |
| 3 | Ṁ14 | |
| 4 | Ṁ5 | |
| 5 | Ṁ4 |
People are also trading
GPT 5.6 sol beats Fable on DeepSWE by 3 points. 5.6 Sol also beats Fable on Agent's Last Exam by 13 points according to the official OpenAI blog post, but falls behind Fable by 1 point on Artificial Analysis Intelligence Index (v4.1). 5.6 Sol beats Fable by 5 points on Terminal-Bench 2.1, but loses to Fable on AA-Briefcase.
I'm ignoring the METR Time Horizon because they don't provide an official score for 5.6 Sol, and don't have any score at all for Fable.
5.6 Sol also loses to Fable on Simplebench.
As of writing, 5.6 Sol is at 1636 while Fable is at 1649 on the WebDev arena on LMArena (https://arena.ai/leaderboard/code/webdev)
After watching Theo t3's video, 5.6 Sol and Fable trade blows and cancel each others wins/losses out. So they're even
Market resolves to No since 5.6 seems to match Fable overall. It wins some and loses some.
5.6 preview is out but the benchmarks aren't out yet
I'll wait for the full release + benchmarks