Qwen3.8-27B-ASCII-Condensed

Qwen3.8-27B with an ASCII-only vocabulary, for long context on a single 16 GB GPU.

Up to 135,168 tokens of context on an RTX 5070 Ti (16 GB).

The same quantization unpruned reaches 114,688 on the same card, so the vocabulary prune is worth about +20,000 tokens (1.18x).

The vocabulary is reduced from 248,320 to 129,006 rows. No weights were retrained, requantized, or otherwise altered — 864 of 866 tensors are bit-identical to the base quant, and the two vocabulary tensors are exact row-subsets of theirs.

What this is

Built from unsloth/Qwen3.8-27B-GGUF UD-IQ4_XS by removing non-ASCII vocabulary rows from token_embd and output, then rewriting the tokenizer to match.

Embeddings are a gather, and the output head is a GEMV over the whole vocabulary on every token. Shrinking both frees VRAM that becomes KV cache, which is what buys the context.

Made with https://github.com/bsaleh03/ASCII-Condensed-prune-tools

What it is not: a finetune, a distillation, or a requantization.

Quickstart

llama-cli -m Qwen3.8-27B-ASCII-Condensed-UD-IQ4_XS.gguf \
  -ngl 99 -fa on -ctk q4_0 -ctv q4_0 \
  -ot "token_embd.weight=CPU" \
  -c 135168 -ub 128 -b 512

-ot "token_embd.weight=CPU" keeps the embedding table in system RAM. It is a pure gather, so this costs ~1% of decode speed and frees its full footprint.

Configuration

Maximum context measured on a 16 GB card (headless), with a real 15K-token prompt — a context that merely allocates is not a usable context.

Throughput

Measured at q4_0/q4_0, -c 135168, on an RTX 5070 Ti:

Decode slows as the KV cache fills — budget for the depth you actually work at, not the shallow-context figure.

* Mixed K/V types need a patched llama.cpp. On stock builds -ctk and -ctv must match — mismatched types silently disable flash attention and cost ~37× prefill speed, with no warning. Use q4_0/q4_0 or q8_0/q8_0 unless you know your build supports mixed types.

MTP (speculative decoding)

The model carries its MTP head, giving ~1.8× decode at reduced context:

--spec-type draft-mtp --spec-draft-n-max 3 -ctkd q4_0 -ctvd q4_0

The draft model keeps its own KV cache and does not inherit -ctk/-ctv — it defaults to f16, so pass -ctkd/-ctvd explicitly.

Language support

The vocabulary is ASCII-only. All 256 byte-fallback tokens are retained, so nothing is unrepresentable — but non-ASCII text costs more tokens:

Nothing breaks; it is a token-efficiency tax proportional to how non-ASCII your text is. If you work in a non-English language, this build is not for you.

Verification

• 864 / 866 tensors byte-identical to the base quant (BLAKE2b compared)
• Surviving vocabulary rows are bit-exact copies — the tables were row-gathered in quantized space, with no dequantize/requantize step
• All 276 special tokens and all 256 byte-fallback tokens retained
• Merge rules filtered so every surviving rule has surviving parents
• bos/eos/pad ids remapped; chat template unchanged
• On ASCII text the tokenizer produces identical output to the base model, token for token

Known limitations

• Text only. No mmproj is provided and the vision pathway is untested.
• Non-ASCII input costs extra tokens (see above).
• Context ceilings are for a 16 GB card. More VRAM shifts them all upward.
• Mixed -ctk/-ctv requires a patched llama.cpp (see Configuration).

Credits and license

• Qwen for Qwen3.8-27B
• unsloth for the GGUF quantization; this build's weights are theirs, unaltered

Apache 2.0, inherited from the base model; LICENSE included.

Modifications from the base: vocabulary reduced 248,320 → 129,006 rows; token_embd.weight and output.weight row-gathered to the retained set; tokenizer.ggml.tokens, token_type, merges and special-token ids rewritten to match. No other tensor was modified.