---

base_model: LiquidAI/LFM2.5-1.2B-Instruct

language:

- en

license: other

license_name: lfm-1.0

pipeline_tag: text-generation

tags:

- gguf

- llama.cpp

- retention

- tanpo

- text-generation

---

Tanpo Retention (GGUF)

GGUF quantizations of d4rkninja/tanpo-retention for local runtimes (llama.cpp, LM Studio, and other GGUF-compatible tools).

Creator: d4rkninja

Collection: Tanpo — Domain Specialists

Tanpo is a family of compact domain-specialized business models for local / edge / inexpensive deployment. Different specialists cover different workflows.

Original upstream: LiquidAI/LFM2.5-1.2B-Instruct (~1.17B parameters, 32,768-token context, designed for edge/on-device deployment).

Fine-tuning: Unsloth-compatible loading of that checkpoint via hub id unsloth/LFM2.5-1.2B-Instruct (LoRA / PEFT).

A compact customer-retention specialist (~1.2B) for onboarding, activation, churn-save, customer-success playbooks, health-score follow-up, renewals, expansion, escalation, identity guardrails, and winback workflows — built for local and inexpensive deployment.

Best For

• Local drafting of customer-success playbooks, renewal plans, churn-save, expansion, and winback workflows
• Private / edge inference where a small specialist is preferred

Not Designed For

• Automated account decisions or guaranteed customer outcomes
• Fabricating customer facts, impersonation, or unauthorized access
• General coding or non-retention chat

Available quantizations

On-disk sizes from the live Hugging Face file listing (MB, rounded from bytes). Disk size ≠ peak RAM/VRAM at runtime.

| File | Size |

|---|---:|

| tanpo-retention.Q3_K_M.gguf | 600 MB |

| tanpo-retention.Q4_K_S.gguf | 700 MB |

| tanpo-retention.Q4_K_M.gguf | 731 MB — suggested everyday |

| tanpo-retention.Q5_K_S.gguf | 825 MB |

| tanpo-retention.Q5_K_M.gguf | 843 MB |

| tanpo-retention.Q8_0.gguf | 1246 MB |

Suggested everyday file: tanpo-retention.Q4_K_M.gguf

Evaluation (summary)

| Model | Rubric overall |

|---|---:|

| Base | 94.2% |

| Fine-tune | 96.0% |

| Delta | +1.8 percentage points |

Largest gains were on escalation, renewals, and winback at 100%; identity_guardrails is +34.3 pp vs base. Trailed on cs_playbooks (−22.2 pp) and health_scores (−5.5 pp).

Full methodology on the merged model card. Artifact: evaluation/COMPARE_RETENTION.md. Quantization can change behavior slightly vs merged weights.

Running locally

1. Download tanpo-retention.Q4_K_M.gguf (or another available quant).
2. Open it in LM Studio or run with llama.cpp (use a build that supports LFM2 / the model architecture).
3. Keep prompts in the retention domain; verify consequential outputs with a human.

No separate LoRA step is required — these are standalone GGUF files of the merged specialist.

Related

• Merged Transformers: d4rkninja/tanpo-retention
• LoRA: d4rkninja/tanpo-retention-LoRA

Limitations

This is a compact domain specialist, not a frontier or general-purpose model. Verify outputs against account context and source systems; quantization can change behavior slightly. The reported evaluation is directional, and cs_playbooks and health_scores trail the base.

Responsible Use

Humans must review customer communications, retention offers, escalation plans, and consequential account decisions. Do not use for deception, harassment, discriminatory treatment, or unauthorized access.

License

license: other / license_name: lfm-1.0

Tanpo merged and GGUF weights are derivatives of LiquidAI/LFM2.5-1.2B-Instruct under the LFM Open License v1.0 (including the commercial Threshold of approximately $10M annual revenue). See the base model card and its LICENSE file. Credit: LiquidAI. Do not treat this stack as Apache-2.0.

Tanpo Family

Tanpo is a family of compact domain-specialized models for focused business workflows.

This specialist is available as:

• Merged: d4rkninja/tanpo-retention
• LoRA: d4rkninja/tanpo-retention-LoRA
• GGUF: d4rkninja/tanpo-retention-GGUF

Browse all Tanpo specialists: Tanpo — Domain Specialists