And with this decision, guys, this day is gonna be remembered as the day of Nvidia's greatest fumble.
1
1
68
If Hydrogen is good and cheap enough, the only thing that would stop google from total domination is their subscription subsidisation.
Quota in Antigravity is so bad subscribing to Google makes no sense
(btw iirk you're forced to share your data just to use Antigravity)
🚨 Gemini 4 Hydrogen has returned 200 on Google Vertex API
We called gemini-4-hydrogen on Vertex AI's global endpoint. It answered "I am Gemini 4 Hydrogen, a large language model built by Google." and the response's model version field says gemini-4-hydrogen.
It isn't in Google's Vertex model catalog yet, and Google hasn't announced it.
2
143
At least now we know what's behind Google's decision to release models three weeks after initial announcement and benchmarks.
3
113
We're saved, they've changed the description of Fable reset!
Yeah, Anthropic are losing concept of good communication again.
What the hell "takes this week’s Fable use off your weekly limit" even mean? I reverts back weekly used for Fable? It discards all the changes made by Fable this week???
1
3
68
"Grok 4.8 will bear no resemblance to 4.7 (thankfully)"
I like it when people from big labs admit that the model was completely disaster.
"Sorry for Opus 5, try this one" energy
We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.
Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.
Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.
1
1
68
I know many of you have been worried about the sudden disappearance of BridgeMind and, more importantly - the lava lamp test...
I had a feeling it should be reintroduced so we can all sleep peacefully knowing which model is currently the best at making lava lamps.
Behold! Today we're comparing our new challenger, Fable 5.2, against its predecessors and OpenAI's current frontier, GPT-6.1 Terra.
One prompt. 4 models. The lava lamp test is NOT dead.
What do you think? Who did the best?
Anthropic gave up first at 'edging customers competition' taking a third place after Google and OpenAI.
I wonder how long before Sol and Astra 6.2
1
1
81
A new way to measure how AI agents handle the routine analyses of wet-lab science—BioBasicsBench.
Developed with Ligo, a biotech company focused on enzyme design, the benchmark draws on 100 tasks from real experiments to test whether agents can interpret raw lab data, identify errors, and recognize when the evidence doesn’t support a conclusion.
Accelerating biological research starts with getting the everyday work right. BioBasicsBench shows where models need to improve, and helps us track what those gains mean for scientific progress, biosafety, and biosecurity.
openai.com/index/measuring-h…
3
79