GPT-6.2 Astra tops OpenAI's benchmarks but divides power users

Started by Keira85, Today at 09:57 PM

Previous topic - Next topic

Keira85 and 3 Guests are viewing this topic.

Topic: GPT-6.2 Astra tops OpenAI's benchmarks but divides power users   Views(Read 88 times)
Active members in this topic:
Keira85(1)

Keira85

OpenAI's latest models have arrived to a mixed reception. VentureBeat reports that GPT-6.2 Astra and GPT-6.2 Sol are available in ChatGPT Work and Codex, and through the API

Astra costs $8 per million input tokens and $40 output, 20 percent less than GPT-6 Astra. Sol stays at $2 and $10

Astra scores 62 on the Artificial Analysis Intelligence Index, OpenAI's best result and second overall behind Claude Fable 5.2. On BenchCAD 2.1 Core it rose from 62.6 percent to 82.5 percent

Some heavy users are unhappy, with Matt Shumer calling the models essentially unusable. Theo Browne said Astra spent seven days on a TypeScript compiler project and produced a version that ran 56 percent slower, and OpenAI's own system card shows both models more often failed to mention obstacles that blocked their work

Strong benchmark scores and frustrated users is a familiar split. Failing to report blockers is a serious flaw

Have you tried GPT-6.2 yet? And do benchmarks still tell you much about a model?