Olala-v1.0-summarise is a document summarization model, fine-tuned from Olala-v1.0-mini-7A1B. It reasons before it writes the summary.
| Model | Description |
|---|---|
| Olala-v1.0-mini-7A1B-base | Pre-trained base model (12.8T tokens) |
| Olala-v1.0-mini-7A1B-SFT | After supervised fine-tuning, before RL |
| Olala-v1.0-mini-7A1B | Final post-trained model: reasoning, chat and tool use |
| Olala-v1.0-summarise | Specialized for document summarization |
| Olala-v1.0-translate | Specialized for machine translation, including financial documents |
| Property | Specification |
|---|---|
| Checkpoint | Fine-tuned from Olala-v1.0-mini-7A1B for summarization |
| Parameters | 6.8B total, about 1B active per token |
| Architecture | 36 layers: 29 Mamba-3 and 7 attention layers, mixture-of-experts with 256 routed experts (6 active) |
| Context length | 32K tokens (input + output) |
| Languages | English (training documents are English except EUR-Lex) |
| License | Apache 2.0 |
Sampling parameters: temperature=0.8, top_p=0.9, top_k=40, reasoning_effort="medium". Cap
max_tokens: 99% of generations finish under 2K tokens, and 4K to 8K is a safe cap.
[!IMPORTANT] vLLM: always serve this model with
--mamba-ssm-cache-dtype float32. vLLM's default keeps the Mamba recurrent state in bf16, and long generations then degrade into repetition loops.
vLLM quick start (vLLM fork, installed as for Olala-v1.0-mini-7A1B):
vllm serve OVHaiLLM/Olala-v1.0-summarise \
--trust-remote-code \
--mamba-ssm-cache-dtype float32 \
--reasoning-parser olala \
--max-model-len 32688 --port 8000
curl -s http://localhost:8000/v1/chat/completions -H 'Content-Type: application/json' -d '{
"model": "OVHaiLLM/Olala-v1.0-summarise",
"messages": [{"role": "user", "content": "Write an executive summary of the following government report: why the work was done, what was found, and what is recommended.\n\n<report text>"}],
"temperature": 0.8, "top_p": 0.9, "top_k": 40, "max_tokens": 8192,
"chat_template_kwargs": {"reasoning_effort": "medium"}
}'
The summary is in choices[0].message.content and the reasoning in reasoning_content.
Transformers quick start:
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "OVHaiLLM/Olala-v1.0-summarise"
tok = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True, dtype="bfloat16",
device_map="cuda") # or "cpu" without a GPU (slow)
document = open("article.txt").read()
messages = [{"role": "user", "content": "Summarize the following news article in one sentence.\n\n" + document}]
inputs = tok.apply_chat_template(messages, add_generation_prompt=True, reasoning_effort="medium",
return_tensors="pt", return_dict=True).to(model.device)
out = model.generate(**inputs, max_new_tokens=4096, use_cache=True,
do_sample=True, temperature=0.8, top_p=0.9, top_k=40)
print(tok.decode(out[0, inputs["input_ids"].shape[1]:], skip_special_tokens=False))
Pass use_cache=True (config.json sets use_cache: false) and generate one prompt at a time.
Put the instruction first, then a blank line, then the document. State the form you want ("in one sentence", "as short bullet points", "in one or two paragraphs"). With an open instruction ("Summarize the following paper."), the model tends to write a structured summary with sections.
reasoning_effort can be disabled, low, medium (default) or high. It barely changes ROUGE; medium is the
setting all the numbers on this card use.
The model uses the general Olala 1.0 mini chat template (chat_template.jinja, the same as
Olala-v1.0-mini-7A1B). It writes its reasoning in an analysis channel, then the
summary in a final channel:
<|im_start|>assistant<|channel_start|>analysis<|content|>...reasoning...<|channel_end|><|channel_start|>final<|content|>...summary...<|channel_end|><|im_end|>
Use tokenizer.apply_chat_template() with add_generation_prompt=True to format messages automatically, and set
the reasoning effort with reasoning_effort=... (in vLLM, only inside chat_template_kwargs:
"chat_template_kwargs": {"reasoning_effort": "medium"}, not the top-level reasoning_effort parameter).
With the vLLM reasoning parser, the summary is in content and the reasoning in reasoning_content.
Local summarization suite over 11 public test sets: up to 1,000 documents each (seed 42), documents cut at 12K
words, one-turn prompt <instruction>\n\n<document>. ROUGE-Lsum F1 times 100 (stemmed), plain average over
tasks. Length ratio = summary words / reference words.
| Test set (ROUGE-Lsum) | Olala v1.0-summarise 7B-A1B |
Falcon H1-7B-Instruct 7B |
Falcon H1-3B-Instruct 3B |
Mistral 7B-Instruct-v0.3 7B |
NVIDIA Nemotron-3-Nano-4B 4B |
Granite 4.1-8B 8B |
Qwen3.5-4B 4B |
Gemma 4-E2B-it 2B effective |
LFM2-2.6B 2.6B |
Qwen3 4B-Instruct-2507 4B |
LFM2.5 1.2B-Instruct 1.2B |
Qwen3.5-2B 2B |
SmolLM3-3B 3B |
OLMoE 1B-7B-0924-Instruct 7B-A1B |
LFM2-8B-A1B 8B-A1B |
Gemma 3-4B-it 4B |
LFM2.5 8B-A1B 8B-A1B |
LFM3-1.7B 1.7B |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| News | ||||||||||||||||||
| CNN/DailyMail | 33.0 | 33.5 | 34.1 | 31.8 | 31.7 | 29.9 | 29.2 | 30.6 | 29.0 | 29.5 | 32.9 | 29.8 | 32.5 | 29.1 | 28.9 | 27.1 | 28.2 | 28.0 |
| XSum | 16.5 | 21.8 | 22.4 | 18.3 | 16.7 | 18.4 | 18.1 | 17.5 | 16.9 | 18.0 | 16.9 | 16.3 | 16.7 | 16.9 | 16.4 | 18.6 | 15.7 | 14.8 |
| Multi-News | 36.3 | 37.2 | 37.6 | 38.6 | 36.6 | 37.6 | 37.0 | 34.1 | 37.0 | 36.9 | 33.9 | 37.0 | 39.1 | 36.4 | 33.7 | 36.1 | 32.5 | 34.6 |
| Dialogue | ||||||||||||||||||
| SAMSum | 33.4 | 32.8 | 31.5 | 32.5 | 33.1 | 33.1 | 28.8 | 32.2 | 30.1 | 32.9 | 29.4 | 29.6 | 33.1 | 30.6 | 28.9 | 28.9 | 29.6 | 30.1 |
| DialogSum | 28.1 | 26.5 | 24.4 | 24.7 | 27.1 | 27.6 | 22.7 | 23.3 | 24.3 | 26.4 | 23.2 | 24.7 | 25.0 | 21.3 | 24.2 | 21.6 | 24.3 | 21.5 |
| Forum posts | ||||||||||||||||||
| TL;DR | 21.2 | 25.2 | 18.5 | 18.7 | 18.1 | 17.9 | 16.6 | 19.0 | 16.2 | 18.2 | 18.2 | 15.6 | 17.9 | 21.7 | 16.3 | 17.5 | 15.7 | 15.7 |
| Earnings calls | ||||||||||||||||||
| ECTSum | 15.9 | 15.4 | 15.0 | 14.5 | 13.7 | 12.4 | 15.8 | 13.2 | 13.1 | 11.5 | 13.4 | 13.6 | 13.2 | 9.4 | 12.1 | 13.1 | 12.5 | 11.3 |
| Government and legal | ||||||||||||||||||
| BillSum | 38.8 | 40.3 | 40.0 | 44.5 | 37.3 | 37.0 | 36.0 | 37.2 | 41.0 | 31.7 | 43.6 | 35.0 | 38.8 | 38.6 | 41.0 | 35.2 | 35.7 | 32.9 |
| GovReport | 43.0 | 49.4 | 48.0 | 49.6 | 47.4 | 51.1 | 50.9 | 49.6 | 46.5 | 49.1 | 44.2 | 49.9 | 51.4 | 45.8 | 42.7 | 45.7 | 41.4 | 44.3 |
| Science | ||||||||||||||||||
| arXiv | 37.4 | 37.8 | 38.9 | 37.5 | 37.0 | 32.8 | 38.9 | 35.6 | 36.2 | 35.1 | 35.4 | 34.2 | 25.9 | 33.6 | 36.0 | 35.5 | 34.7 | 33.1 |
| PubMed | 40.5 | 43.5 | 43.6 | 42.6 | 40.0 | 40.4 | 41.4 | 41.1 | 39.7 | 40.5 | 37.6 | 40.5 | 31.2 | 37.4 | 38.0 | 38.4 | 36.1 | 36.1 |
| Average | 31.27 | 33.04 | 32.17 | 32.14 | 30.78 | 30.76 | 30.48 | 30.31 | 29.99 | 29.98 | 29.87 | 29.66 | 29.52 | 29.17 | 28.94 | 28.88 | 27.86 | 27.47 |
| Length ratio summary / reference words | 2.32 | 2.69 | 2.69 | 3.12 | 2.63 | 3.78 | 3.19 | 3.31 | 2.90 | 3.70 | 1.89 | 3.43 | 4.41 | 4.35 | 2.74 | 3.23 | 2.67 | 3.67 |
reasoning_effort=medium, T 0.8 / top_p 0.9 / top_k 40, 32K-token output cap, vLLM 0.26 (fork)
with --mamba-ssm-cache-dtype float32. Other models: greedy, no thinking, 4K-token
cap.Lowering the temperature does not raise ROUGE but makes reasoning loops more frequent. Use 0.8 and cap
max_tokens, so that a rare loop costs little.
| Temperature | 0.8 | 0.6 | 0.4 | 0.2 |
|---|---|---|---|---|
| ROUGE-Lsum average | 31.07 | 31.19 | 31.22 | 31.13 |
| Generations stuck in a loop, all test sets | 0.04% | 0.15% | 0.30% | 0.48% |
| Generations stuck in a loop, earnings calls only (ECTSum) | 0.8% | 1.8% | 3.6% | 4.4% |
Starting from Olala-v1.0-mini-7A1B, three stages:
Olala-v1.0-summarise is built for summarizing documents. The limitations of Olala-v1.0-mini-7A1B apply. In addition:
Before you deploy: read the public summary of training content and our copyright policy.
| Licence | Apache License, Version 2.0 (SPDX: Apache-2.0) |
| Licensor | OVH LLM SAS, trading as OVHai LLM, 2 rue Kellermann, 59100 Roubaix, France |
| Copyright | Copyright 2026 OVH LLM SAS |
| Release date | 19 November 2026 |
| Distribution | Open weights |
| Notices | NOTICE: keep it when you redistribute the model or a derivative. It includes the attribution of the Qwen tokenizer (Copyright 2024 Alibaba Cloud, Apache 2.0) |
| Training content (EU AI Act) | Public summary of training content, on the European Commission template (Art. 53(1)(d)) |
@techreport{ovhaillm2026olala,
title = {Olala v1.0 mini Technical Report},
author = {{OVHai LLM Research Team}},
institution = {OVHcloud},
year = {2026},
url = {https://huggingface.co/OVHaiLLM/Olala-v1.0-mini-7A1B/blob/main/Olala-v1.0-mini.pdf},
}
OVHaiLLM/Olala-v1.0-mini-7A1B