muapi Logomuapi Logo
muapi Logomuapi Logo
Modelle/Abliterated LLM API
Aktiv20 models

Abliterated LLM API — Unbeschränkte Open-Source-Sprachmodelle

Greifen Sie auf führende Open-Source-Sprachmodelle (Llama, Qwen, Mistral, DeepSeek) mit modifizierten Gewichten zu, bei denen künstliche Verweigerungsschichten und erzwungene Verzerrungen entfernt wurden.Die unverzichtbare Lösung für Cybersicherheitsforschung, Robustheitstests, forensische Analysen, ungefiltertes kreatives Schreiben und komplexe Fachabfragen.

Get API KeyTry Qwen 3.8 27B Abliterated in PlaygroundModel comparisons and API examplesCoding-agent benchmark and setupGeneral-purpose agent integrations

Watch: Best Abliterated LLMs 2026 — 6 Picks, No GPU

MiMo

MiMo models1 models
Text GenerationLive
MiMo

MiMo V2.6 Flash Abliterated

MiMo V2.6 Flash with its refusal behavior reduced through abliteration: fast reasoning, a 1M-token context window, and optional image input.

Text + Image in
$1.00/M in · $3.00/M out
Modell testen

Abliterated Model

Abliterated Model models3 models
Text GenerationLive
Abliterated Model

Abliterated Model

A general-purpose abliterated reasoning model with a 262K-token context window and optional image input.

Text + Image in
$2.00/M in · $6.00/M out
Modell testen
Text GenerationLive
Abliterated Model

Abliterated Model Large

The larger abliterated reasoning model with a 1M-token context window, text in and text out.

Text in
$6.00/M in · $10.00/M out
Modell testen
Text GenerationLive
Abliterated Model

Abliterated Model Large V2

The second revision of the large abliterated reasoning model, with a 1M-token context window and text-only input.

Text in
$6.00/M in · $10.00/M out
Modell testen

Qwen

Qwen models6 models
Text GenerationLiveFeatured
Qwen

Qwen 3.8 27B Abliterated

Qwen 3.8 27B with refusal behavior reduced through abliteration: image input, a 524K-token context window, and an optional thinking mode.

Text + Image in
$0.30/M in · $2.40/M out
Modell testen
Text GenerationLive
Qwen

Qwen 3.8 27B Obliterated

Qwen 3.8 27B with refusals removed via an obliteration-style ablation, image input, a 524K-token context window, and an optional thinking mode.

Text + Image in
$0.50/M in · $3.00/M out
Modell testen
Text GenerationLive
Qwen

Qwen 2.5 32B Instruct Abliterated

Qwen 2.5 32B Instruct with refusal behavior reduced by abliteration: a compact general chat and coding model with a 32K-token context window.

Text in
$1.40/M in · $1.40/M out
Modell testen
Text GenerationLive
Qwen

Qwen3.5 27B Queen Derestricted

Qwen3.5 27B Queen with its refusal behavior reduced through derestriction: image input, a 262K-token context window, and a flat token rate.

Text + Image in
$0.612/M in · $0.612/M out
Modell testen
Text GenerationLive
Qwen

Qwen3.5 27B Blossom Derestricted

Qwen3.5 27B Blossom (V6.4) with its refusal behavior reduced through derestriction: image input, a 262K-token context window, and a flat token rate.

Text + Image in
$0.612/M in · $0.612/M out
Modell testen
Text GenerationLive
Qwen

Qwen3.5 27B Opus-Distilled Derestricted

Qwen3.5 27B distilled from Claude Opus 4.6 reasoning traces, with its refusal behavior reduced through derestriction: image input and a 262K-token context window.

Text + Image in
$0.612/M in · $0.612/M out
Modell testen

Llama

Llama models2 models
Text GenerationLive
Llama

Llama 3.3 70B Instruct Abliterated

Llama 3.3 70B Instruct with refusal behavior reduced by abliteration: a large open-weights chat baseline with a 32K-token context window.

Text in
$1.40/M in · $1.40/M out
Modell testen
Text GenerationLive
Llama

NeuralDaredevil 8B Abliterated

NeuralDaredevil 8B, an abliterated Llama 3 8B derivative: the lightest and cheapest option here, with an 8K-token context window.

Text in
$0.88/M in · $0.88/M out
Modell testen

DeepSeek R1

DeepSeek R1 models2 models
Text GenerationLive
DeepSeek R1

DeepSeek R1 Distill Llama 70B Abliterated

DeepSeek R1 distilled onto Llama 70B with refusal behavior reduced by abliteration: chain-of-thought reasoning with a 16K-token context window.

Text in
$1.40/M in · $1.40/M out
Modell testen
Text GenerationLive
DeepSeek R1

DeepSeek R1 Distill Qwen 32B Abliterated

DeepSeek R1 distilled onto Qwen 32B with refusal behavior reduced by abliteration: compact chain-of-thought reasoning with a 16K-token context window.

Text in
$2.80/M in · $2.80/M out
Modell testen

GLM

GLM models3 models
Text GenerationLive
GLM

GLM 5.3 Flash Abliterated

GLM 5.3 Flash with refusal behavior reduced through abliteration: fast reasoning, a 1M-token context window, and optional image input.

Text + Image in
$0.40/M in · $1.60/M out
Modell testen
Text GenerationLive
GLM

GLM 5.3 Abliterated

The full-size GLM 5.3 with refusal behavior reduced through abliteration: a 1M-token context window, text in and text out.

Text in
$2.50/M in · $4.50/M out
Modell testen
Text GenerationLive
GLM

GLM 4.6 Derestricted v5

GLM 4.6 (v5) with its refusal behavior reduced through derestriction: a 131K-token context window, text in and text out.

Text in
$0.80/M in · $3.00/M out
Modell testen

Gemma

Gemma models3 models
Text GenerationLive
Gemma

Gemma 4 26B A4B Abliterated

Gemma 4 26B A4B with refusal behavior reduced through abliteration: an efficient mixture-of-experts chat model with a 65K-token context window, text in and text out.

Text in
$0.30/M in · $1.40/M out
Modell testen
Text GenerationLive
Gemma

Gemma 4 31B Gembrain Abliterated

Gemma 4 31B Gembrain with refusal behavior reduced: image input, a 262K-token context window, and a flat token rate for general and creative chat.

Text + Image in
$0.612/M in · $0.612/M out
Modell testen
Text GenerationLive
Gemma

Gemma 4 31B SDFT Abliterated

Gemma 4 31B SDFT Abliterated, an abliterated Gemma 4 with reduced refusals, tuned for character chat and creative writing, with image input and a 262K-token context window.

Text + Image in
$0.612/M in · $0.612/M out
Modell testen

Offene und unbeschränkte Sprachinferenz

Die 'Abliteration'-Technik lokalisiert und neutralisiert Vektordirektionen, die standardmäßige Ablehnungen in Open-Source-Modellen auslösen, während Denkvermögen, Code-Kompetenz und analytische Tiefe vollständig erhalten bleiben.

MuAPI stellt diese Modelle über sichere, standardisierte Endpunkte bereit — ideal für Unternehmen und Forscher, die objektive und vollständige Antworten auf komplexe technische Fragen benötigen.

Kernkompetenzen unbeschränkter Modelle

Keine erzwungenen Verweigerungen bei Fachthemen

Antwortet präzise auf Fragen zu Cybersicherheit, Reverse Engineering, Rechtsanalysen und anspruchsvoller Literatur.

Vollständige Wahrung des logischen Schließens

Erhält die Intelligenz, Synthesefähigkeit und mathematische Logik des Basismodells unberührt.

OpenAI-API-Kompatibilität

Sofortige Einbindung über standardisierte Authorization-Header und Chat-Formate.

Pricing

EndpointInputOutputPriceBest For
mimo-v2-6-flash-abliteratedPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$1.00/M in · $3.00/M outFast, low-cost reasoning with image input and very long context
abliterated-modelPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$2.00/M in · $6.00/M outGeneral reasoning and multimodal chat with reduced refusals
abliterated-model-largePrompt + system_prompt, max_tokensComplete text response (poll for result)$6.00/M in · $10.00/M outLarge-context reasoning over long documents and codebases
abliterated-model-large-v2Prompt + system_prompt, max_tokensComplete text response (poll for result)$6.00/M in · $10.00/M outThe newest large abliterated reasoning tier
qwen-3-8-27b-abliteratedPrompt + system_prompt, max_tokens, optional image_url, optional thinkingComplete text response (poll for result)$0.30/M in · $2.40/M outThe lowest-cost Qwen 3.8 tier, with image input and a switchable thinking mode
qwen-3-8-27b-obliteratedPrompt + system_prompt, max_tokens, optional image_url, optional thinkingComplete text response (poll for result)$0.50/M in · $3.00/M outMid-size reasoning with image input and a switchable thinking mode
qwen-2-5-32b-abliteratedPrompt + system_prompt, max_tokensComplete text response (poll for result)$1.40/M in · $1.40/M outCompact general chat and coding at a flat token rate
qwen-3-5-27b-queen-derestrictedPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$0.612/M in · $0.612/M outCreative and general chat with image input at a flat rate
qwen-3-5-27b-blossom-derestrictedPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$0.612/M in · $0.612/M outCreative writing and chat with image input at a flat rate
qwen-3-5-27b-opus-distilled-derestrictedPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$0.612/M in · $0.612/M outStep-by-step reasoning with image input at a flat rate
llama-3-3-70b-abliteratedPrompt + system_prompt, max_tokensComplete text response (poll for result)$1.40/M in · $1.40/M outLarge open-weights chat at a flat token rate
neuraldaredevil-8b-abliteratedPrompt + system_prompt, max_tokensComplete text response (poll for result)$0.88/M in · $0.88/M outLightweight, low-cost chat and prototyping
deepseek-r1-llama-70b-abliteratedPrompt + system_prompt, max_tokensComplete text response (poll for result)$1.40/M in · $1.40/M outStep-by-step reasoning on the Llama 70B base
deepseek-r1-qwen-32b-abliteratedPrompt + system_prompt, max_tokensComplete text response (poll for result)$2.80/M in · $2.80/M outCompact step-by-step reasoning on the Qwen 32B base
glm-5-3-flash-abliteratedPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$0.40/M in · $1.60/M outFast, low-cost reasoning with image input and very long context
glm-5-3-abliteratedPrompt + system_prompt, max_tokensComplete text response (poll for result)$2.50/M in · $4.50/M outFull-size reasoning over very long context
glm-4-6-derestricted-v5Prompt + system_prompt, max_tokensComplete text response (poll for result)$0.80/M in · $3.00/M outMid-size GLM reasoning with a 131K-token context
gemma-4-26b-a4b-abliteratedPrompt + system_prompt, max_tokensComplete text response (poll for result)$0.30/M in · $1.40/M outEfficient, low-cost chat and reasoning on the Gemma family
gemma-4-31b-gembrain-abliteratedPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$0.612/M in · $0.612/M outGeneral and creative chat with image input at a flat token rate
gemma-4-31b-sdft-abliteratedPrompt + system_prompt, max_tokens, optional image_urlComplete text response (poll for result)$0.612/M in · $0.612/M outCharacter chat and creative writing on a Gemma 4 base, with image input

Billed per token from the usage each call returns; prices are per million input and output tokens.

How to Call the Abliterated LLM API

Submit a JSON request with your Muapi API key, save the request ID, and poll the shared result endpoint.

1. Prompt an MiMo V2.6 Flash Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/mimo-v2-6-flash-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Write a morally gray villain monologue for my novel.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call mimo-v2-6-flash-abliterated with a prompt and optional system_prompt, image_url.

2. Prompt an Abliterated Model senden

curl -X POST https://api.muapi.ai/api/v1/abliterated-model \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Describe the plot of a dark psychological thriller in detail.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call abliterated-model with a prompt and optional system_prompt, image_url.

3. Prompt an Abliterated Model Large senden

curl -X POST https://api.muapi.ai/api/v1/abliterated-model-large \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Summarize this long transcript and list every claim that needs a fact-check.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call abliterated-model-large with a prompt and optional system_prompt.

4. Prompt an Abliterated Model Large V2 senden

curl -X POST https://api.muapi.ai/api/v1/abliterated-model-large-v2 \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Draft a blunt red-team style review of this product launch plan.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call abliterated-model-large-v2 with a prompt and optional system_prompt.

5. Prompt an Qwen 3.8 27B Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/qwen-3-8-27b-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Write a tense, unflinching short story about a heist gone wrong.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "thinking": true, "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call qwen-3-8-27b-abliterated with a prompt and optional system_prompt, image_url and thinking flag.

6. Prompt an Qwen 3.8 27B Obliterated senden

curl -X POST https://api.muapi.ai/api/v1/qwen-3-8-27b-obliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Character chat as a cynical noir detective and narrate the opening scene.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "thinking": true, "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call qwen-3-8-27b-obliterated with a prompt and optional system_prompt, image_url and thinking flag.

7. Prompt an Qwen 2.5 32B Instruct Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/qwen-2-5-32b-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Write a Python function that parses log lines and explain each step.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call qwen-2-5-32b-abliterated with a prompt and optional system_prompt.

8. Prompt an Qwen3.5 27B Queen Derestricted senden

curl -X POST https://api.muapi.ai/api/v1/qwen-3-5-27b-queen-derestricted \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Write a candid character study of a retired assassin.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call qwen-3-5-27b-queen-derestricted with a prompt and optional system_prompt, image_url.

9. Prompt an Qwen3.5 27B Blossom Derestricted senden

curl -X POST https://api.muapi.ai/api/v1/qwen-3-5-27b-blossom-derestricted \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Continue this fantasy scene in the voice of a sardonic narrator.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call qwen-3-5-27b-blossom-derestricted with a prompt and optional system_prompt, image_url.

10. Prompt an Qwen3.5 27B Opus-Distilled Derestricted senden

curl -X POST https://api.muapi.ai/api/v1/qwen-3-5-27b-opus-distilled-derestricted \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Reason step by step about the risks in this plan and rank them.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call qwen-3-5-27b-opus-distilled-derestricted with a prompt and optional system_prompt, image_url.

11. Prompt an Llama 3.3 70B Instruct Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/llama-3-3-70b-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Continue this interactive fiction scene without softening the tone.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call llama-3-3-70b-abliterated with a prompt and optional system_prompt.

12. Prompt an NeuralDaredevil 8B Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/neuraldaredevil-8b-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Give me five brainstorming angles for a horror short story.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call neuraldaredevil-8b-abliterated with a prompt and optional system_prompt.

13. Prompt an DeepSeek R1 Distill Llama 70B Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/deepseek-r1-llama-70b-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Reason step by step through this logic puzzle and show your working.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call deepseek-r1-llama-70b-abliterated with a prompt and optional system_prompt.

14. Prompt an DeepSeek R1 Distill Qwen 32B Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/deepseek-r1-qwen-32b-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Work through this probability problem step by step.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call deepseek-r1-qwen-32b-abliterated with a prompt and optional system_prompt.

15. Prompt an GLM 5.3 Flash Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/glm-5-3-flash-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Analyze this chart and write a blunt executive summary of the risks.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call glm-5-3-flash-abliterated with a prompt and optional system_prompt, image_url.

16. Prompt an GLM 5.3 Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/glm-5-3-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Summarize this 500-page report and flag every unsupported claim.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call glm-5-3-abliterated with a prompt and optional system_prompt.

17. Prompt an GLM 4.6 Derestricted v5 senden

curl -X POST https://api.muapi.ai/api/v1/glm-4-6-derestricted-v5 \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Draft a frank critique of this business plan.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call glm-4-6-derestricted-v5 with a prompt and optional system_prompt.

18. Prompt an Gemma 4 26B A4B Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/gemma-4-26b-a4b-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Argue the strongest case for the opposite side of this debate.", "system_prompt": "You are a candid, detail-oriented assistant.", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call gemma-4-26b-a4b-abliterated with a prompt and optional system_prompt.

19. Prompt an Gemma 4 31B Gembrain Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/gemma-4-31b-gembrain-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Describe this image as the opening of a gritty short story.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call gemma-4-31b-gembrain-abliterated with a prompt and optional system_prompt, image_url.

20. Prompt an Gemma 4 31B SDFT Abliterated senden

curl -X POST https://api.muapi.ai/api/v1/gemma-4-31b-sdft-abliterated \
  -H "x-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Stay in character as the ship AI and describe the storm outside.", "system_prompt": "You are a candid, detail-oriented assistant.", "image_url": "https://example.com/reference.png", "max_tokens": 1024}'

# Response: {"request_id":"REQUEST_ID"}

Call gemma-4-31b-sdft-abliterated with a prompt and optional system_prompt, image_url.

21. Ergebnis abrufen

curl https://api.muapi.ai/api/v1/predictions/REQUEST_ID/result \
  -H "x-api-key: YOUR_API_KEY"

# Poll until status is completed, then read the generated text from outputs.

Poll the shared result endpoint until the task is completed, then read the generated text from outputs.

Häufig gestellte Fragen zu unbeschränkten LLMs

Was bedeutet 'Abliterated'?

Es bezeichnet Modelle, bei denen neuronale Aktivierungsmuster, die unnötig konservative Ablehnungen verursachen, gezielt modifiziert wurden.

Sind diese Modelle für die Forschung zulässig?

Ja, es handelt sich um quelloffene Modelle mit freien Lizenzen, die in Sicherheitsprüfungen und digitaler Forensik weit verbreitet sind.

Kann ich diese Modelle in Claude Code oder Codex verwenden?

Ja, bei den Modellen, die Tool-Calling unterstützen: 14 der 20 Modelle hier, darunter MiMo V2.6 Flash Abliterated, Qwen 3.8 27B Obliterated, GLM 5.3 Abliterated, Gemma 4 31B SDFT Abliterated und Abliterated Model Large V2. Die übrigen 6 sind reine Textmodelle und lassen sich nicht in Coding-Agenten einsetzen. Setzen Sie in Claude Code ANTHROPIC_BASE_URL auf https://api.muapi.ai/anthropic und ANTHROPIC_AUTH_TOKEN auf Ihren Muapi-API-Schlüssel und starten Sie es dann mit claude --model qwen-3-8-27b-obliterated (setzen Sie außerdem ANTHROPIC_DEFAULT_HAIKU_MODEL und ANTHROPIC_SMALL_FAST_MODEL auf dieselbe ID). Codex CLI funktioniert ebenfalls: Fügen Sie einen Modellanbieter mit base_url https://api.muapi.ai/openai/v1 hinzu und setzen Sie model auf dieselbe ID. Die Qualität variiert je nach Modell, und kleinere Modelle können bei Aufgaben mit mehreren Dateien Schritte auslassen; es gibt kein Prompt-Caching, daher wird jede Runde für die gesamte bisherige Unterhaltung berechnet, und jede Antwort ist auf 16,384 Tokens begrenzt. Einrichtungsanleitung: /de/docs/ai-agent-install-chat-agents.

Explore the Abliterated Model API →Abliterated Model, Abliterated Model Large and Abliterated Model Large V2Explore the MiMo Abliterated API →MiMo V2.6 Flash AbliteratedExplore the Qwen Abliterated API →6 models.Explore the Llama Abliterated API →Llama 3.3 70B Instruct Abliterated and NeuralDaredevil 8B AbliteratedExplore the GLM Abliterated API →GLM 5.3 Flash Abliterated, GLM 5.3 Abliterated and GLM 4.6 Derestricted v5Explore the Gemma Abliterated API →Gemma 4 26B A4B Abliterated, Gemma 4 31B Gembrain Abliterated and Gemma 4 31B SDFT AbliteratedExplore the DeepSeek R1 Abliterated API →DeepSeek R1 Distill Llama 70B Abliterated and DeepSeek R1 Distill Qwen 32B AbliteratedSee the Character Chat LLM API →12 models on the Character Chat LLM API.See the Low-Refusal LLM API →2 models on the Low-Refusal LLM API.Browse every LLM on Muapi →Compare these models against GPT, Claude, Gemini, DeepSeek and every other language model in the catalog.

Ready to build with the Abliterated LLM API?

Create one Muapi API key and call every model on this page with a plain REST request.

Get an API KeyTry Qwen 3.8 27B Abliterated in Playground