Part of Engrym Seed — Orvyth's seed-tier local model family for tool-using agents. Qwen3.5
hybrid linear-attention architecture, 262,144-token native context, first-class tool calling.
The strongest stable artifact. Substantially slower per tool call — prefer Base or Flash for agent loops.
Weights are distributed via the Ollama registry.
ollama run Orvyth/engrym-seed:pro
Size
28.6 GB
77-task score
134/143 (93.7%)
Tool calling
12/12 verified
Context
262,144 native (default num_ctx 32,768)
Quantization
Q8_0
The full ladder
Tag
Class
Size
77-task
:nano
Nano 2B
2.1 GB
90.8/143
:flash
Flash 4B
4.6 GB
124/143
:base
Base 9B
9.5 GB
131/143
:pro-27b-q4
Pro 27B v2 Q4
16.5 GB
134/143
:pro
Pro 27B v2 Q8
28.6 GB
134/143
:pro-e
Pro-E 27B (experimental)
28.6 GB
137/143
Evaluation
77 tasks · 143 points · temperature=0 · max_tokens=16384 · seed=42 · one attempt ·
deterministic validators · no LLM judge. Scores are bound to the exact published blobs.
Asking the model to work deliberately recovers points on tasks it otherwise fails, and the gain is
largest for the smallest models (Nano +11.8, Flash +6, Base +3, Pro-E +0). On the small end that is
worth more than a model upgrade.
If a prompt exceeds num_ctx, Ollama returns HTTP 400 — it does not silently truncate.
Lineage
Qwen/Qwen3.5 base → Ornith-1.0-9B x Qwythos-9B TIES merge (9B line) → Orvyth identity and
chip-calling LoRA merged into the weights → converted and quantized in-house.
Limits
First-party scores; treat small gaps between adjacent models as unresolved.
The identity tune is light; a heavy external system prompt can pull behavior toward the base model.
The 27B is much slower per tool call than the 9B and smaller.
No MTP speculative-decoding head in these builds.
Tags are mutable — pin the digest for production and evaluations.
ORVYTH - Intelligence. Governed. Ground truth over hype. Prove before you claim.
