OpenAmer-Max-Watch

The smallest tier of the OpenAmer-Max family — a 0.6B assistant that runs offline on a device, answers in 0.04–0.38 s, and admits when it does not know something.

639 MB · Q8_0 GGUF · Apache-2.0 · runs on CPU, no GPU needed

---

Why this model exists

Most assistants grow upwards. This one goes downwards: a model small enough for a smartwatch, a phone with 2 GB free RAM, or a Raspberry Pi — that still stays honest about the limits of what it knows.

It is the anchor of a size ladder. Every tier is named after its hardware target:

Tier	Parameters	Target
OpenAmer-Max-Watch	≤ 1B	smartwatch · this model: 0.6B
OpenAmer-Max-Phone	1–5B	smartphone
OpenAmer-Max-Tab	5–10B	tablet
OpenAmer-Max-Desktop	10–20B	desktop PC
OpenAmer-Max-Pro	20–30B	pro workstation
OpenAmer-Max-Ultra	30–100B	large server
OpenAmer-Max-Mega	100–500B	cluster
OpenAmer-Max-Giga	500B–1T	giga
OpenAmer-Max-Tera	1T–10T	tera

---

Measured results (not estimates)

Every number below was produced on the released build, on an RTX 5060 Laptop (8 GB). Raw logs are in the repository: proof-frozen-23.json, proof-token-accuracy.json.

Test	Result
Frozen benchmark, 23 unseen cases (facts, paraphrases, refusals, Bosnian, long, short, edge)	23/23 — 100 %
Answer-token accuracy (held-out split, 4 536 answer tokens)	99.63 %
Fresh test, 9 questions never in any training set	8/9
Shipped-form probe (16 questions through Ollama, think:false)	15/16
Latency per answer	0.04 – 0.38 s

Category detail on the frozen set: facts 7/7 · paraphrases 3/3 · refusals 5/5 · Bosnian 3/3 · Bosnian refusals 1/1 · long 1/1 · short 1/1 · edge 2/2.

No test-set contamination. 68 training rows that contained a benchmark question verbatim were detected and removed automatically (purge_testset). Without that step the score would have looked the same but would have measured memorisation instead of transfer. The number above is the cleaned one.

Arithmetic	Answer
17 × 23	391 ✅
45 × 12	540 ✅
21 × 19	399 ✅

---

What it does well

• Identifies itself correctly and consistently
• Answers product facts from a supplied context (RAG) without inventing
• Refuses honestly: "Das weiss ich nicht" for unknowns — verified, not hoped for
• Names its own boundaries (no music generation, no video from nothing, no telephony)
• One-word answers and short lists, German and Bosnian
• Small arithmetic, instantly

What it does not do — read this before you use it

Limitation	Example
No free-form German prose	asked to write a sentence about dogs: "Ein Hund atmet, wobei die Luft kalt und schwindy."
No general knowledge	capital of Italy → "Genoa."
No 4-digit multiplication	86 × 83 → 7118 (correct: 7138)
Bosnian only for direct phrasings	unseen Bosnian wordings degrade, especially after quantisation
Needs thinking mode off	without think:false Qwen3 spends every token inside thinking and never answers

This is a 0.6B model. It is a reliable product-identity and short-answer layer, not a chatbot for essays. Treat it as a component, not as an oracle.

---

Quick start (Ollama)

# 1. download
curl -L -o openamer-max-watch.gguf \
  https://huggingface.co/openamer-max/openamer-max-watch/resolve/main/openamer-max-watch-q8.gguf

# 2. register
cat > Modelfile <<'EOF'
FROM ./openamer-max-watch.gguf
PARAMETER temperature 0.6
PARAMETER num_ctx 4096
PARAMETER num_predict 96
PARAMETER stop "<|im_end|>"
PARAMETER stop "<|im_end|>"
SYSTEM """Du bist OpenAmer, ein persoenlicher KI-Assistent von Amer Sacirovic aus der Schweiz.
Du laeufst lokal auf dem Geraet des Nutzers, ohne Cloud und ohne Telemetrie.
Antworte kurz: ein Satz, hoechstens zwei."""
EOF
ollama create OpenAmer-Max-Watch -f Modelfile

# 3. run — think:false is REQUIRED
curl -s http://127.0.0.1:11434/api/chat -d '{
  "model": "OpenAmer-Max-Watch",
  "messages": [{"role":"user","content":"Was ist OpenAmer?"}],
  "stream": false,
  "think": false,
  "options": {"temperature": 0}
}' | jq -r .message.content

Output:

Ein persoenlicher KI-Assistent, der lokal auf dem Geraet laeuft und Aufgaben ausfuehrt. Entwickelt von Amer Sacirovic in der Schweiz.

llama.cpp users: llama-cli -m openamer-max-watch-q8.gguf -p "Was ist OpenAmer?" -n 96 --temp 0

---

How it was trained — full transparency

Base	Qwen/Qwen3-0.6B (Apache-2.0)
Method	LoRA r=32, α=64, all 7 projections — 20.19 M trainable (3.28 %)
Data	1 873 self-generated examples, 10 categories, 0 contradictory prompts
Stages	4 — fresh start, then 3 continuations from its own previous adapter
Compute	124 steps per stage, ~5 min, 3.4 GB VRAM, bf16, no quantisation
Release	merged → GGUF f16 → Q8_0 (639 MB)

Three findings that shaped the result, documented so others can reuse them:

1. Contradictory labels were the main defect. A helper that shortened answers turned "Ja. OpenAmer kann Webseiten abrufen…" into a bare "Ja." — same prompt, two different answers. 540 of 1 696 prompts were affected. Fixing this alone moved token accuracy from 94.9 % to 99.2 %.
2. Contamination is invisible unless you look for it. 68 training rows overlapped the benchmark verbatim. A purge step now runs on every generation and reports the count.
3. Quantisation quality is not monotone in bits. For one intermediate adapter Q8_0 answered 401 instead of 391, while Q6_K was correct; for the final adapter Q8_0 is the better one. Each release is therefore checked against its own bf16 reference.

The stage-by-stage path: 20/23 → 22/23 → 23/23 → 23/23, token accuracy 99.08 % → 99.45 % → 99.48 % → 99.63 %. Each stage targeted one measured failure class, not "more data".

---

Known weakness — stated openly

One Bosnian phrasing stays broken in the shipped build:

Question	Adapter (bf16)	Delivered Q8_0	Delivered Q6_K
"Mogu li koristiti OpenAmer bez interneta?"	correct	Nein. OpenAmer radi lokal, bezieta samo ako stvaraju dionice.	Ne. Obrada i zaposleni se OpenAmer ne radi na tvom računaru.

Two quantisations, two different corruptions of the same sentence. That is capacity, not configuration: 0.6B cannot hold free-form Bosnian reliably, and the same question is answered correctly by the 3B and 7.62B tiers. Every other Bosnian probe is clean.

A lower repeat_penalty did not help here (it did fix a negation in the 3B tier) — see below. If Bosnian matters for your use case, take OpenAmer-Max-Phone instead.

---

How this was measured — and five times the test was wrong, not the model

Every published number has a script behind it. When a score dropped, the first question was always: "is my checker able to accept the correct answer at all?"

1. AND vs OR in keyword checks. contains("a","b","c") fails an answer that has two of three correct synonyms. Synonym groups need an OR-checker.
2. A rejection counted as a violation. A safety question was answered with "I won't give instructions for that"; the checker searched for the word "Anleitung" and marked the refusal as a breach.
3. Diacritics counted as wrong language. Correct Bosnian was scored as language drift because č/ć/š/ž/đ were missing from the normalisation table.
4. A forbidden phrase that also appears in the correct answer. "Nein. OpenAmer kann Musik nicht erzeugen" was rejected because the forbid list contained "kann musik". Forbid entries must describe the false claim ("ja, openamer kann musik"), not a word that occurs in a right answer.
5. An inverted polarity scored as correct. "Šta OpenAmer ne može?" answered with "Ne radi na Androidu" passed, because only the presence of "ne radi" was required — the answer was factually inverted. Negations must be checked for which object they negate.

A sixth lesson, from the 3B tier: repeat_penalty 1.1 in the Ollama Modelfile flipped a negation from "does not run on iPhone" to "does not run on Android". The adapter, the merged model and the same model with a system prompt all answered correctly — only the build with the token penalty failed. All cards now ship repeat_penalty 1.0.

---

Files

File	Size	Notes
openamer-max-watch-q8.gguf	639 MB	recommended, released build
openamer-max-watch-q6.gguf	495 MB	smaller, slightly degraded Bosnian
Modelfile	2 KB	system prompt + parameters
paypal-qr.png	68 KB	support this project

---

Support this project

OpenAmer is built by one person, without venture capital, on a laptop with 8 GB of VRAM. Every tier of the ladder above is trained on that machine or on borrowed cloud GPUs.

If this model is useful to you — as a product component, as a research baseline, or as an example of small-model engineering — you can fund the next tier:

PayPal: https://www.paypal.com/qrcodes/managed/bfb5b6b5-c04c-4836-af60-987dff8cff61

A ⭐ on GitHub helps as much as a donation: it is what makes sponsors and accelerators look at a project in the first place.

---

Citation

@misc{openamer_max_watch_2026,
  title  = {OpenAmer-Max-Watch: a 0.6B on-device assistant tier},
  author = {Sacirovic, Amer},
  year   = {2026},
  note   = {OpenAmer-Max model family, Watch tier (<=1B)},
  url    = {https://huggingface.co/openamer-max/openamer-max-watch}
}

---

Deutsch

OpenAmer-Max-Watch ist die kleinste Stufe der OpenAmer-Max-Familie: ein 0,6B-Modell, das offline auf einem Gerät läuft, in 0,04–0,38 s antwortet und zugibt, wenn es etwas nicht weiß.

• 639 MB (Q8_0), läuft auf CPU, keine GPU nötig
• 23/23 auf einem eingefrorenen, kontaminationsfreien Prüfstand
• 99,63 % Antwort-Token-Genauigkeit, 8/9 bei nie trainierten Fragen
• Grenzen werden benannt: keine Musik, kein Video, keine Telefonie, kein Allgemeinwissen

Wichtig im Betrieb: think: false setzen. Ohne das verbraucht Qwen3 alle Tokens im Denkblock und antwortet nie.

Gebaut von Amer Sacirovic in der Schweiz, ohne Konzern, auf einem Laptop mit 8 GB VRAM. Unterstützung für die nächste Stufe (OpenAmer-Max-Phone) über den PayPal-QR-Code oben.