Japan's National Institute of Informatics released LLM-jp-4-33b-thinking — a 33B dense, Apache 2.0 reasoning model, one of three variants in the LLM-jp-4 series alongside an 8B dense and a 32B-A3B MoE (128 routed experts, 8 active, 3.8B active parameters). All three share a 65,536-token context and a training run of 11.7 trillion tokens across pretraining and mid-training. What makes it worth a post isn't the size — 33B is modest by this month's standards — it's who built it and how.
A public research institute, not a company
NII isn't a lab racing for market share; it's Japan's national institute for informatics research, and LLM-jp is closer to public infrastructure than a product line. That shows up in the release itself: pretraining, mid-training, SFT, and DPO datasets are all publicly linked from the model card, with only the portions genuinely blocked by third-party licensing withheld. That's a rare level of pipeline transparency — the same standard we credited Soofi S for out of Germany, now matched from Japan's academic sector rather than a startup. It also lands as a direct counterpoint to Sakana AI's Namazu, Japan's other answer to sovereign AI, which explicitly skips pretraining and post-trains other labs' open weights instead. NII is the "actually train one from scratch" half of the same national conversation — proof the cheaper strategy isn't the only one Japan is running.
The methodology choice worth naming: reasoning without RL
The model card states it plainly, almost in passing: post-training uses supervised fine-tuning and direct preference optimization, without reinforcement learning. That's the detail worth sitting with. Nearly every "thinking" or reasoning-capable release this month — Qwen3.8-Max, DeepSeek-V4-Pro, GLM-5.3 — has built its reasoning behavior through an RL stage, several of them shipping mandatory always-on reasoning with no way to disable it as a direct consequence. LLM-jp-4-33b-thinking gets a working reasoning mode from SFT and DPO alone. Whether that's a resource constraint, a methodological preference, or a genuine finding that RL isn't necessary for reasoning behavior at this capability tier, the card doesn't say — but it's a clean natural experiment sitting right next to a month of RL-heavy releases, and worth watching for anyone trying to isolate what RL is actually buying labs that use it.
The benchmark table, read directly
NII evaluates with an LLM-as-judge setup using gpt-5.4-2026-03-05 — and states outright that switching judges from the gpt-4o used for the prior llm-jp-3 generation makes scores stricter and not directly comparable across generations, the kind of methodology disclosure worth crediting on its own. Against that judge, on MT-Bench (JA/EN), AnswerCarefully (a 336-question Japanese safety benchmark), and llm-jp-instructions (400 single-turn questions): LLM-jp-4-33b-thinking beats gpt-oss-20b — OpenAI's own open-weight release — on every metric at comparable reasoning effort settings: MT-Bench Japanese 8.00 vs. 7.33, English 8.24 vs. 7.85, AnswerCarefully 3.79 vs. 3.55, llm-jp-instructions 3.79 vs. 3.16. Against gpt-5.4 itself, the closed flagship, the gap stays wide and honest — 8.98 vs. 8.00 on Japanese MT-Bench, 4.41 vs. 3.79 on safety. A national research model beating an American lab's own open release on Japanese-language tasks, while trailing that same lab's closed frontier model by a clear margin, is a specific and legible result rather than a vague "competitive" claim.
What's still unstated
The card is explicit about being early-stage: "not tuned to ensure outputs align with human intent and safety considerations" beyond the disclosed benchmarks. There's no comparison against other sovereign or Japanese-market models — not Namazu, not Sakana's other releases, not Rakuten's or other domestic efforts — so the model's position relative to Japan's own AI ecosystem, rather than against OpenAI's models, is still an open question. And the chat template's Harmony-format compatibility comes with a real caveat: the tokenizer differs from what the openai-harmony library expects, so direct tokenization with that library isn't supported — a specific compatibility gap for anyone assuming drop-in interoperability. The companion cookbook repository makes the practical shape of that gap concrete: running these models means either enabling trust_remote_code — the models bundle custom tokenizer and Harmony-parsing plugins to work correctly at all — or manually integrating that same code from the cookbook and the separate llm-jp-tokenizer project by hand. That's a real deployment friction point for anyone in a trust_remote_code-averse environment, and it's not mentioned on the model card itself.
What to expect next
- Watch for independent benchmark reproduction, particularly the gpt-oss-20b comparison — it's the most checkable claim on the card and the one most relevant to anyone choosing between open Japanese-language models.
- Watch whether the no-RL approach gets revisited in a future generation. If NII adds an RL stage to LLM-jp-5 and reasoning scores jump, that's real evidence about what RL contributes; if they don't, that's evidence too.
- Watch how this sits alongside Namazu. Two different Japanese sovereignty strategies — train from scratch versus post-train someone else's weights — are now both live and both published with real numbers, a genuine natural comparison this dataset can track over time.
References: Hugging Face — llm-jp/llm-jp-4-33b-thinking · GitHub — llm-jp/llm-jp-4-cookbook · related coverage: Japan's Sovereign AI Strategy Is to Not Train a Model · Soofi S: Europe's Sovereign Open-Weight Foundation Model · The Most Openly Licensed Frontier Model Is Korean · GLM-5.3 · Frontier Arcade: trends & predictions