Translation

Cursorbench: Grok 4.5 better than GPT-5.5, at ~half the cost

Cursorbench, a new coding benchmark, ranks Grok 4.5 above GPT-5.5 in performance while costing roughly half as much, according to results posted on cursor.com.

Background

Cursor is a popular AI-powered code editor built on VS Code. They published a benchmark comparing AI models on a set of software engineering tasks. - "Grok" refers to models made by xAI (Elon Musk's company). Grok 4.5 is their latest generation. - "GPT" refers to models made by OpenAI. There is no official "GPT-5.5" — this is Cursor's internal labeling for a version of GPT-5 that they tested. - Cursor's claim is that Grok 4.5 outperforms GPT-5.5 on coding tasks while costing roughly half as much per use. - This matters because developers and companies choose which AI model to use based on a tradeoff between capability and cost. If xAI's model is both better and cheaper, it could shift market share away from OpenAI.

RT Max Weinbach: I had early access to this model and I'll post more Thursday, but I think the most impressive part is this model NEVER GIVES UP If yo...
9.0
Max Weinbach says he had early access to OpenAI's new model GPT-5.6 Sol, calling it his favorite model by far. He highlights that it never gives up and will keep reasoning until it's done. OpenAI announced that GPT-5.6 Sol, along with Terra and Luna, will launch publicly on Thursday, with preview access expanding globally now.
Better Models: Worse Tools
7.5
Newer Claude models sometimes invent extra keys in tool call arguments, breaking validation in Pi's edit tool. The author suspects post-training for Claude Code's forgiving harness makes alternative schemas fail. This suggests closed RL training can degrade general tool-use reliability.
Writing an LLM from scratch, part 34b -- from bigrams to GPT-2, one component at a time (in JAX)
3.0
The author builds and trains a GPT-2 small model from scratch in JAX, starting from a basic bigram-style model and incrementally adding components like LayerNorm and Transformer blocks. Achieved a final loss of 3.418, beating their PyTorch version (3.538) and original GPT-2 small (3.499) on the same test dataset.
Flax debugging: making a hash of things
2.0
The author debugged a Flax NNX training loop where the loss was stuck at 10.82, indicating random guessing. By hashing the model parameters and comparing hashes across steps, they discovered the parameters weren't changing. The root cause was using @jax.jit instead of @nnx.jit, which is needed for proper in-place state propagation of parameter updates in NNX.
Writing an LLM from scratch, part 34a -- building a JAX training loop for an LLM training run
1.0
Building a JAX training loop for an LLM from scratch using Flax NNX and Optax. First validated the harness with a minimal "A-to-A" model (embedding then projection) before adding Transformer layers. Key challenges included fixing JAX's GPU memory default and slow data iteration by committing data to CPU memory.

Cursorbench: Grok 4.5 better than GPT-5.5, at ~half the cost

Background

Related stories

RT Max Weinbach: I had early access to this model and I'll post more Thursday, but I think the most impressive part is this model NEVER GIVES UP If yo...

Better Models: Worse Tools

Writing an LLM from scratch, part 34b -- from bigrams to GPT-2, one component at a time (in JAX)

Flax debugging: making a hash of things

Writing an LLM from scratch, part 34a -- building a JAX training loop for an LLM training run