Clawthor
LongformLongWriter-6k dataset

Longform Writer

longform-v1

Research papers, reports, documentation, articles — fine-tuned for 10,000+ word outputs

Technical Specs

Architecture

Llama 3.1 8B (LongWriter)

Context Window

32,768 tokens

Max Output

16,384 tokens

Streaming

Yes

Custom System Prompts

Yes

Category

Longform

Why This Model

THUDM’s LongWriter, a Llama 3.1 8B model fine-tuned on the LongWriter-6k dataset — 6,000 examples specifically designed for generating outputs exceeding 10,000 words in a single generation. Most language models hit a wall around 2,000 words regardless of context window size. LongWriter was trained to break through that wall. With vLLM, it produces 10,000+ words within a minute. 32K context window for ingesting long source material. Built for the use cases other models abandon: full research papers, comprehensive reports, complete documentation sets.

Default System Prompt

Minimal by design — the model's fine-tuning does the heavy lifting. Override with a custom system parameter if needed.

Write extended documents. Structured, thorough, well-organized. Produce complete, detailed output.

API Usage

curl -X POST https://clawthor.xyz/api/v1/generate \
  -H "Authorization: Bearer ck_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "longform-v1",
    "prompt": "Write a comprehensive research report on the state of L2 scaling in 2026",
    "max_tokens": 500
  }'

Resources