The Kaitchup – AI on a Budget
Subscribe
Sign in
Home
Notes
AI Notebooks
The Kaitchup's Book
Weekly Kaitchup
Tutorials
The Kaitchup Index
Archive
About
Latest
Top
Discussions
Gemma 3 270M: Can Tiny Models Learn New Tasks?
A case study with machine translation
Sep 1
•
Benjamin Marie
13
10
August 2025
The Uneven Impact of Post-Training Quantization in Machine Translation
The Weekly Kaitchup #107
Aug 29
•
Benjamin Marie
2
NVFP4: Same Accuracy with 2.3x Higher Throughput for 4-Bit LLMs
How to quantize LLMs with NVFP4
Aug 25
•
Benjamin Marie
8
2
DeepSeek V3.1 and a New Nemotron Hybrid Model
The Weekly Kaitchup #106
Aug 22
•
Benjamin Marie
3
Qwen3 Instruct "Thinks": When Token Budgets Silently Skew Benchmark Scores
Replicating Qwen3-30B-A3B-Instruct-2507 shows accuracy riding on verbosity, not just GSPO
Aug 18
•
Benjamin Marie
3
2
GPT-OSS 120B Tops The Kaitchup Index
The Weekly Kaitchup #105
Aug 15
•
Benjamin Marie
3
How to Run Unsloth on Multi-GPU Setups: Data-Parallel or Model-Parallel
Step-by-step fixes for running Unsloth across GPUs
Aug 11
•
Benjamin Marie
6
3
OpenAI GPT-OSS: Native 4-Bit MoE Models
Everything you need to know about GPT-OSS 20B and 120B, MXFP4 quantization, and running on Blackwell & H100 GPUs -- The Weekly Kaitchup #104 Special…
Aug 7
•
Benjamin Marie
11
4
GSPO vs GRPO: Reinforcement Learning for MoE Models
How Qwen’s GSPO Outperforms GRPO for Stable and Scalable MoE Training
Aug 4
•
Benjamin Marie
4
The Weekly Kaitchup #103
Qwen3-2507 - DailyBench - Gemma 3 QAT Evaluation
Aug 1
•
Benjamin Marie
9
July 2025
H100 PCIe vs SXM vs NVL: Which H100 GPU Is Fastest and Most Cost-Effective for Fine-Tuning LLMs?
Benchmarking all three H100 variants for full, LoRA, and QLoRA fine-tuning
Jul 28
•
Benjamin Marie
9
The Weekly Kaitchup #102
Qwen3’s “Dehybridization” - Qwen3-Coder - The Kaitchup Index
Jul 25
•
Benjamin Marie
6
1
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
