Funny, I was testing the Horizon Alpha model on @FeatureCrewPod
's maze tool. In the regular chat mode it was always just responding instantly without thinking and failing. But when I clicked the 'app' button on open router, it took time to think and create an app that actually
@petergostev
-
Horizon Alpha Model Shows Better Performance With Extended Thinking
By
–
-
Claude Model Cannibalization: Token Usage Trends Sonnet Releases
By
–
How much are Claude models cannibalising themselves? After each new release, usage of the previous model declines, yet the overall consumption increase, e.g., weekly @openrouter usage rose from 211b tokens for Sonnet 3.7 to 287b tokens for Sonnet 4 (a one-third increase).
-
AI Model Performance Assessment on Coding Tasks
By
–
This is quite good though But tested on some coding tasks, didn't seem that much better
-
OpenAI Horizon Alpha Expected Release Date August 2026
By
–
Previously when OpenAI tested models (Quasar Alpha & Optimus Alpha) it took 4-12 days to release the model afterwards. So if we extrapolate this for Horizon Alpha, the dates are 3-11 August
-
Stargate Definition OpenAI Long Term Cloud Contracts
By
–
So looks like the definition of Stargate is 'any [new?] clusters that OpenAI has long term contract with?'
-
Horizon Alpha Model Performance: Quick Review and Comparison
By
–
Horizon Alpha – not bad honestly, for such a quick model – much better than Nectarine / Starfish; worse than Lobster. But better than most existing models at this task, e.g. better than Kimi K2, or GPT-4.1 https://t.co/NrLFTgkze6 pic.twitter.com/blK1GPoITJ
— Peter Gostev (@petergostev) 31 juillet 2025Horizon Alpha – not bad honestly, for such a quick model – much better than Nectarine / Starfish; worse than Lobster. But better than most existing models at this task, e.g. better than Kimi K2, or GPT-4.1
-
OpenAI’s General Approach to IMO Gold Medal Achievement
By
–
The most interesting part about OpenAI getting a gold medal in the IMO competition is not really the result itself but rather how it was achieved. Two things were interesting:
— Peter Gostev (@petergostev) 30 juillet 2025
1) OpenAI are deliberately prioritising methods that are general rather than specific;
2) They have… pic.twitter.com/yPsGNyvTUIThe most interesting part about OpenAI getting a gold medal in the IMO competition is not really the result itself but rather how it was achieved. Two things were interesting: 1) OpenAI are deliberately prioritising methods that are general rather than specific; 2) They have
-
GPT-5 Code Capabilities Threaten Anthropic Revenue
By
–
What happens to that revenue if GPT-5 is much better at code? Cursor et al swith over & Anthropic revenue down 50% overnight?
-
OpenAI Major Release Strategy Beyond Google Tokyo Event
By
–
Yeah but feels like a random Tokyo event isn't meant to be for big Google announcements. Hard to imagine that OpenAI would try to tie their biggest release to Google this time round. A demo of a random impressive looking feature (e.g. the voice mode) was one thing, but this, I
-
GPT-4o Thinking Mode Triggered Across All Model Variants
By
–
So it looks like it triggers GPT-4o with thinking, regardless of the model – it does the same thing (thinking for 35 seconds) regardless whether I select 4o, o3, or o3-pro