Alpha
Contact
Use · Tucano 2

Is it good enough?

Ownership levelPartialnone·limited·partial·substantial·fullAnalytical input D ยท 54.4/100

This page is a projection of the one entry record, the Reliability factor that Use covers. The full verdict is set by all four factors together, floor-weighted so the weakest caps the whole.

Which domain expands which factor
  • AssessUse & modify + Transparency
  • ImplementData control + Reliability
  • UseReliability
  • SupportTransparency

Capabilities & modalities

Text only. The post-training data covers 12 task types: code, function calling, general instructions, maths, maths with chain of thought, reasoning, retrieval-augmented answers, rewriting, structured output, summarisation, system prompts and translation (ev-gv2-post). The 3.7B Instruct leads its 3 to 4B class on Knowledge and Reasoning (56.22) and trails Qwen3-4B on instruction following and coding (ev-paper). The Think models write chain-of-thought in Portuguese inside think tags and have no coding training. All figures are the publisher's own.

Context window & long-context behaviour

The context is 4,096 tokens in config.json and tokenizer_config.json. Qwen3-4B-Base supports 32,768, and the Tucano continual pretraining and post-training stop at 4,096 (ev-card-instruct, ev-qwen3-licence). RULER-PT at 4,096 tokens is 0.686 for the 3.7B Instruct and 0.707 for the 3.7B Think, against 0.979 for Qwen3-4B. The paper attributes the gap to missing long-context data and lists context extension as future work (ev-paper). Chunk documents and retrieve.

Prompt format & chat template

The format is ChatML. Turns open with <|im_start|> and the role, and close with <|im_end|>. A tools block lists function signatures in <tools></tools> in Portuguese and asks for a JSON call inside <tool_call></tool_call> (ev-configs). The Think template wraps reasoning in <think> and </think>. The tokenizer is a 49,152-token Portuguese vocabulary with <|im_end|> as the end-of-sequence token. Use the repo's apply_chat_template. The Base models have no chat format.

Language coverage

Portuguese only. The cards say other languages may cause misinterpretation, and the continual pretraining data is purely Portuguese (ev-card-instruct, ev-paper). Qwen3 covers 119 languages, and the Tucano tokenizer swap and Portuguese training reduce that coverage in ways the publisher does not measure. The Instruct card shows an English translation task.

Function / tool calling

The Instruct SFT mix includes about 17.5 million tokens of function-call data, and the card shows a correct create_todo call in <tool_call> format (ev-configs, ev-card-instruct). No tool-calling benchmark is reported, so test your own schemas.

Structured / JSON-constrained output

The Instruct SFT mix includes about 35 million tokens of structured-output data, and the card shows a JSON extraction with the requested keys (ev-configs, ev-card-instruct). No benchmark or schema-guarantee is reported. Use constrained decoding in your serving layer if you need valid JSON every time.

How this scores

The ownership factor this domain covers, drawn from the one entry record.

3

ReliabilityIs it reliable and good enough for the job?

Weak

Safety is rated 2 because no safety evaluation, acceptable-use policy or guard model is published (ev-card-instruct, ev-paper). Operational readiness is 3, with no official quantisation or serving guide.

How this scores (AOI sub-dimensions)
Operational3/5how practical it is to run, serve and maintain in productionThe weights use the standard Qwen3 architecture in transformers with safetensors.
Safety2/5whether misuse risks are evaluated and guardrails are providedThe Instruct and Think models are safety-tuned with a published 28,437-pair preference set built from written constitutions, and the cards list hallucination, bias, toxicity and repetition.
What this means for adoptionYou own the weights: Apache-2.0, ungated, commercial use allowed, with the code and data to retrain the Tucano stages on your own hardware. You do not own a verified safe assistant, because no safety evaluation exists, and you cannot rebuild the Qwen3 stage underneath. The level rises if Polygl0t publishes a safety evaluation and resolves the mixed licence of the training corpus.

Sources

The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.

Model cardread2026-11-19
The 3.7B Instruct card declares Apache-2.0, 3,759,341,056 parameters and a 4,096-token context, and describes one SFT round and one APO round on gigaverbo-v2-sft and gigaverbo-v2-preferences.
Model cardread2026-11-19
The 3.7B Base card says the model is continually pretrained from Qwen3-4B-Base on about 50 billion tokens, with the Tucano2 tokenizer transplanted by Orthogonal Matching Pursuit.
Model cardread2026-11-19
The 3.7B Think card declares Apache-2.0 and describes a reasoning model trained with SFT and APO that writes chain-of-thought inside think tags.
Model cardread2026-11-19
The Tucano2-0.6B-Base card declares Apache-2.0, a Llama architecture with 670,127,616 parameters, about 400 billion training tokens and three published stage configs.
Licenceread2026-11-19
Eight of the nine Qwen-derived repos and the 0.6B Base repo carry an identical Apache-2.0 LICENSE file that names four individuals as copyright holders.
Licenceread2026-11-19
Qwen3-4B-Base, 1.7B-Base and 0.6B-Base each carry an Apache-2.0 LICENSE file and are ungated.
Licenceread2026-11-19
Qwen2.5-7B, 14B and 32B-Instruct, which generated the synthetic data, and the abliterated 32B variant that wrote the rejected responses, are Apache-2.0 with LICENSE files and ungated.
Technical_reportread2026-11-19
The paper 'Tucano 2 Cool' (arXiv 2603.03543, 3 March 2026) describes continual pretraining from Qwen3 bases, a from-scratch 0.6B model, the data, the post-training recipe and the publisher's benchmarks.
Documentationread2026-11-19
The GigaVerbo-v2 card lists its source datasets with their licences and states that the combined dataset has no single unified licence (license: other).
Documentationread2026-11-19
The Synth, SFT and Preferences cards declare Apache-2.0 and are ungated.
Documentationread2026-11-19
llm-foundry (training, data, alignment and evaluation code) is Apache-2.0 and the Portuguese lm-evaluation-harness fork is MIT.
Documentationread2026-11-19
The Polygl0t Hugging Face organisation is not verified and has seven members.
Documentationread2026-11-19
The Instruct and Think repos publish SFT and APO YAML configs, training logs, config.json and a ChatML chat template.
Third-party analysisread2026-11-19
A third-party repo ran the 3.7B Instruct in bf16 on vLLM 0.33.0 on an RTX 3080 with 8.7 GiB of VRAM and scored 74.6% on 393 text-only ENEM questions.