Amazon Fine‑Tunes GPT‑OSS‑20B for Multilingual Reasoning on SageMaker HyperPod
Current overview (not historical)
Amazon’s SageMaker HyperPod platform was used to fine‑tune the open‑weight GPT‑OSS‑20B model for multilingual reasoning. The base model, which only reasoned in English, achieved target‑language chain‑of‑thought (CoT) in just 22 % of 50 multilingual CoT tests. A two‑stage supervised fine‑tuning (SFT) process first taught the model to reason in the target language after about 200 steps (CoT accuracy 100 %) and then, after roughly 1,000 steps, to provide answers in that language. Gradient‑Reinforced Policy Optimization (GRPO) was applied to preserve the 100 % CoT accuracy while boosting the target‑language answer accuracy by an additional 8 %. The results demonstrate that with a modest amount of fine‑tuning, an English‑only foundation model can be adapted to perform robust reasoning and answer generation across multiple languages on Amazon’s scalable infrastructure.
Written with AI from the reporting · updated 10/20/2026
Expand the detailed fact route and sources
Story route
A station-by-station route from the first public evidence onward; every station carries its own evidence and uncertainty.
- 1Unclassified fact stationOct 21, 00:00recorded Oct 21, 02:36single source
Fine-tuning GPT-OSS-20B for multilingual reasoning on Amazon
Read original source ↗ AWS Machine Learning BlogReplay this moment in the explorerEvidence & revisions · 1 source records
- Subject：
- openai/gpt-oss-20B
- Action：
- fine-tuned for multilingual reasoning
- Object：
- GPT-OSS-20B model
- Occurred (fact side)：
- Oct 21, 00:00
- Published (valid side)：
- Oct 21, 01:48
- First seen (known side)：
- Oct 21, 02:36
Fact stationSingle source · one-sided claimCopy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.
Uncertainty marks flag where to check the sources; they are not proven causation.
Report timeline
Explore every side through dated source reports.
- AWS Machine Learning BlogFirst-party sourceRelated fact: single sourceAWS fine-tunes GPT-OSS-20B for multilingual reasoning on SageMaker HyperPod
Report first recorded Oct 21, 02:36
GPT-OSS-20B, an open-weight reasoning model, defaults to English chain-of-thought despite user instructions to reason in French, German, Spanish, or Italian. On a 50-example multilingual test, the base model reasoned in the correct target language only 22 percent of the time, nearly all from English prompts. AWS researchers fine-tuned the model using supervised fine-tuning (SFT) and reinforcement learning with Group Relative Policy Optimization (GRPO) on Amazon SageMaker HyperPod. SFT alone was often sufficient to establish the target language behavior, while GRPO provided additional refinement when the model already had some capability. A controlled experiment with Qwen2.5-7B confirmed that matching the technique to the specific capability gap determines the outcome.
Read original source ↗
Event discussion
Reader opinions are hosted on GitHub Discussions; posting requires a GitHub account. Comments are not verified facts.