Weaver: Claude-Style Verbosity for Roleplay and Narrative Text
Weaver (alpha) is a text-only model from Mancer, listed as a LLaMA 2 13B model for creative, verbose output. It was released in August 2023 and sits in Mancer's roleplay-oriented lineup alongside MythoMax and ReMM-SLerp. Its stated aim is to recreate Claude-style verbosity for roleplay and narrative use, with a warning that coherence and memory will not match Claude. It fits character dialogue, scene-setting and story generation, not reasoning, tool use or document-heavy work.
Add Weaver to your narrative pipeline via AnyAPI.ai
Llama 2 13B alpha tuned for verbose roleplay and narrative prose, with 8K context and up to 6K output tokens.
Modality:
Text
model ID
mancer/weaver
Output Speed *
N/A
tok/s
Intelligence Index *
N/A
/ 100
Context Window *
8000
tokens
Input price
2.4
Anytoken
Output price
4.5
Anytoken
Performance
Built for Long, Descriptive Output, Not for Coherence or Recall
Weaver's strength is output length rather than accuracy. Mancer lists it as a creative, verbose 13B model with an 8,000-token context window and a 6,000-token output ceiling, so one reply can use most of the window. The provider's own description warns that coherence and memory will trail Claude. That makes it a fit for scene-setting and story openings, where length and tone matter more than consistency. For long sessions, re-inject character sheets and recaps. No independent benchmark data was found for this model.
Technical Specifications
What the model supports
Weaver accepts and returns text only, with an 8,000-token context window and a 6,000-token maximum output. Treat the two limits separately: a long prompt leaves less room for generation inside the window, and a maximum-length reply consumes most of it. The listed endpoint does not accept tools, so function calling is unavailable. It is an alpha-labelled 13B model with no documented reasoning mode, so use it for text generation, not agent loops or structured pipelines.
Verified Specifications —
Weaver
*
Input modalities
Text
output modalities
Text
Context window
8000
tokens
Maximum output tokens
2000
Reasoning
No
Knowledge cutoff
August 2023
Pricing (standard)
2.4
AnyTokens in
/
4.5
AnyTokens out
Limitations & Trade-offs
1
Weak coherence and memory. Mancer's own description says Weaver attempts Claude-style verbosity but should not be expected to match Claude's coherence or memory. In long roleplay sessions, characters may drift or forget earlier facts. Compensate with recaps and re-injected character sheets, or choose a larger, newer model when continuity across many turns is critical.
2
No tools, reasoning or verified structured output. The listed endpoint does not accept tools, there is no documented reasoning mode, and structured outputs are not advertised. Weaver cannot drive agent loops, function-calling workflows or schema-bound extraction. Use a tool-capable model for those steps and keep Weaver for the creative text layer.
3
Small 8,000-token window. Eight thousand tokens covers a modest conversation or a few scenes, but not long documents, large lorebooks or retrieval-heavy prompts. A long prompt also reduces room for the 6,000-token maximum output. For long-running campaigns, RAG, or document processing, a model with a substantially larger context window is the better choice.
4
Alpha status and thin evidence. Weaver carries an alpha label, dates from August 2023, and has no independent benchmark data. Its listing does not document the training approach beyond the LLaMA 2 13B architecture. Speed figures also differ between sources, so measure latency and throughput yourself before putting it in a latency-sensitive product.
Best-Fit Workloads
Where this model earns its place
01
In-character roleplay dialogue
Weaver is positioned by Mancer for roleplay and narrative situations, and its verbose style suits persona-driven chat where replies should be rich and descriptive. It works well for companion-style apps and character simulations driven by a system prompt. Because memory is explicitly flagged as weaker than Claude's, pair it with session summaries or periodic persona reminders.02
Story openings and scene-setting
The 6,000-token output ceiling lets Weaver produce long passages in a single response, which suits story openings, scene descriptions and atmospheric prose. Verbosity is its stated design goal. Use it to draft or expand narrative text that a human or a stronger model will later edit for consistency and fact-checking.03
Interactive fiction prototyping
Text-only input and output, an 8,000-token window and adjustable sampling parameters make Weaver a lightweight option for prototyping branching stories or game dialogue. Keep game state, rules and inventory outside the model, since it has no tool or structured-output support. Move to a larger-context model if campaigns need long persistent memory.04
Integration
01
One API integration
02
Easy model switching
03
Flexible for production
04