[mchinen]

The frontier graph on all these benchmark are extremely in favor of 5.6 Sol over Fable, more than the best model comparisons in previous iterations.

I'd like to know how cherry-picked this is, and what tests it performed less overwhelmingly in, but I suppose that info is not going to be on this post.

If it pans out to be as good as it says, that's great. On the other hand, if this model is not overwhelmingly impressive over Fable, I will lose what remaining trust I had in these announcements.

[thurn]

They do disclose that they scored much lower than Fable on SWEBench Pro, which is a pretty high-quality benchmark. I think it's partially just about what they choose to emphasize...

[danielsamuels]

It's worth noting that OpenAI recently came out saying, "We don't think SWEBench Pro is worth reporting any more" - https://openai.com/index/separating-signal-from-noise-coding...

[bel8]

SWEBench Pro should be ignored until they fix it or disprove the broken task accusations.

[futureshock]

There has been a lot of chatter ever since the Mythos scores had been release that SWEbench pro had major contamination and that Mythos had memorized many questions that lacked the context to be solvable on their own. And now with OpenAI saying a large number of the questions are broken, I think it's worth taking that single outlier benchmark with some salt when the overall trend is that 5.6 is very competitive with Mythos at about half the price.

[simianwords]

> SWEBench Pro, which is a pretty high-quality benchmark

No, doesn't seem like it

https://openai.com/index/separating-signal-from-noise-coding...

[mchinen]

I totally missed that, because in the charts they showcase for coding, the SWEBench score is not present, they only include it at the end of the post in tables. Hmm.

Great catch.

[saberience]

The SWEBench benchmarks are really gamed at this point and should not be trusted period. The solutions are effectively in the training sets and have been for a while.

[mnicky]

SWE Bench Pro is completely different benchmark than SWE Bench (e.g. Verified) suite was. It only copied the name.

[saberience]

SWE Bench Pro is also gamed and shouldn't be trusted.

[mnicky]

Then we are left with what? FrontierCode maybe? IIRC that one evaluates not only if tests pass but also code quality - e.g. whether the maintainer would accept the pull request as is.

[Sabinus]

Why that benchmark in particular?

[lwansbrough]

Didn't they also just post about how SWEBench is broken?

[enraged_camel]

The charts are also extremely difficult to parse. They seem auto-generated. Dataset coloring is atrocious.

Regarding your main point, yes, I agree. My impression (as someone who uses both Codex and Claude Code daily) is that OpenAI does a fair amount of benchmaxxing.

[therobots927]

The proof is in the pudding and these benchmark stats will only work for so long before people lose interest.