Model library / FINAL-Bench
Darwin-27B-RSI is a 27-billion-parameter Qwen-3.5 dense language model that has been recursively self-improved using only its own generated solutions.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Darwin-27B-RSI is a dense Qwen-3.5-27B model released by FINAL-Bench. The publisher describes it as a thinking model designed for graduate-level science reasoning, trained without any human-written answers and evaluated on benchmarks such as GPQA Diamond, GSM8K and MMLU-Pro. It is intended to be used with generous token budgets to allow extensive reasoning.
The model contains 26.9 billion parameters, uses a context window of 262,144 tokens, and is provided in bfloat16 weight format. It is licensed under Apache-2.0 and has no expert layers. Deployment should respect the licence, allocate sufficient memory for the 27 B dense architecture, and can follow the recommended sampling settings of temperature 0.6 and top_p 0.95 for hard problems.
Specification
| Repository | FINAL-Bench/Darwin-27B-RSI |
| Publisher | FINAL-Bench |
| Published | 27 September 2026 |
| Architecture | Qwen3_5ForCausalLM |
| Model type | qwen3_5_text |
| Parameters | 26,895,998,464 |
| Layers | 64 |
| Hidden size | 5,120 |
| Attention heads | 24, 4 key/value heads |
| Context length | 262,144 tokens |
| Vocabulary | 248,320 tokens |
| Weight format | bfloat16 |
| Licence | apache-2.0 |
| Base model | FINAL-Bench/Darwin-27B-Opus |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 50.1 GiB | Full precision as released |
| fp8 | 25.0 GiB | 8-bit, near-lossless on most models |
| int4 | 12.5 GiB | 4-bit, smallest footprint |
The cache costs 256 KiB per token at 16-bit, so the full 262,144-token context of one request needs about 64.0 GiB. Concurrency multiplies that number, not the weights.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 4 | 26.3 GiB | about 107,734 tokens |
| RTX 5090 32GB | 2 | 2.5 GiB | about 10,249 tokens |
| L40S 48GB | 2 | 31.3 GiB | about 128,214 tokens |
| A100 80GB | 1 | 19.4 GiB | about 79,471 tokens |
| H100 80GB | 1 | 19.4 GiB | about 79,471 tokens |
| RTX PRO 6000 96GB | 1 | 33.8 GiB | about 138,454 tokens |
| H200 141GB | 1 | 74.3 GiB | the full 262,144 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve FINAL-Bench/Darwin-27B-RSI \ --max-model-len 106496 \ --gpu-memory-utilization 0.90
Questions
The weights take about 50 GiB at the released precision. That fits on one A100 80GB, which leaves roughly 19 GiB for the key-value cache.
The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 262,144 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
Hy-Memory-30B-A3B is a Tencent-developed long-term-memory writer model designed to replace LLMs in Mem0, MemOS and Hy-Memory fram…
Granite Switch 4.1 30B is a modular LLM combining a 30-billion base, decision and guardian adapters, and an English speech-to-tex…
Sabi-Noema-MLA is a 27.4 B-parameter agentic model derived from Qwen3.8-27B, using multi-head latent attention to reduce attentio…
LFM2.5-24B-A2B is a 24 billion-parameter MoE reasoning model with 2.45 billion active parameters for on-device agentic tasks.
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.