MagmaRouter vs self-hosted LiteLLM
LiteLLM Proxy is the build-it-yourself version of what MagmaRouter sells: an open-source, OpenAI-compatible gateway with virtual keys and budgets. If you want full control, self-hosting wins. What it costs you is the ops and the provider keys.
Facts checked September 2026 against self-hosted LiteLLM’s own site, docs and policies.
| MagmaRouter | self-hosted LiteLLM | |
|---|---|---|
| What it is | Hosted gateway for one uncensored, tool capable model line (Xploded). | An open-source (MIT) OpenAI-compatible proxy you run yourself, in front of your own provider keys, with virtual keys and budgets. |
| Pricing | Pay as you go, $5 minimum top-up, no subscription required. | Free and open source. You pay the model providers directly at their list price, with nothing added by a gateway. Some enterprise features (SSO, audit logs) are paid. |
| OpenAI-compatible | Yes | Yes |
| Anthropic-compatible | Yes, /v1/messages (verified with Claude Code) | Yes, an Anthropic-format pass-through is available |
| Uncensored / abliterated | Seven uncensored models, the Xploded line: GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash 1017, DeepSeek V4.1 Flash Lite 1017, MiMo V2.6 Flash, Qwen3.8 27B and Qwen3.8 27B Ultrafast. Every one calls tools. | None by itself: you point it at whatever providers you configure, so the uncensored catalogue is on you to source |
| Prompt retention | No prompt or completion stored; model, tokens and cost kept 90 days; signed prompt-hash receipt | Your server, your rules: logging is configurable and off unless you wire a logging callback; nothing leaves your control except to the providers you choose |
| Per-key spend caps | Lifetime and monthly per-key caps, enforced by reserving the worst case and clamping max_tokens | Per-virtual-key max_budget with a budget_duration reset window, plus rate limits; worst-case budget reservation is on by default |
Where self-hosted LiteLLM is the better choice
- Full control and privacy: it is free, and prompts go only to the providers you configure, from a server you control. Nothing passes through a third-party gateway.
- Total control and real budgets: your own key policies, routing, logging and model list, with virtual-key budgets that reserve the worst case before the call, on by default. Its spend controls are as capable as MagmaRouter’s or more.
Where MagmaRouter is different
- You run it: hosting, upgrades, uptime, secrets and the provider contracts are all yours, and you must source uncensored models yourself. MagmaRouter is that work already done, with the uncensored catalogue included.
- Both reserve the worst-case cost, but MagmaRouter shrinks max_tokens to fit the cap instead of rejecting the request, does it as a hosted service, and signs a prompt-hash receipt on every call. LiteLLM gives you the mechanism; MagmaRouter gives you the running system.
The uncensored models MagmaRouter sells
If you are comparing self-hosted LiteLLM because you want an uncensored or abliterated model API, these are the models on MagmaRouter: GLM 5.3 uncensored, GLM 5.3 Flash uncensored, DeepSeek V4.1 Flash 1017 abliterated, DeepSeek Flash Lite abliterated, MiMo V2.6 Flash abliterated, Qwen3.8 27B uncensored and Qwen3.8 27B Ultrafast uncensored. Every one calls tools, works in Claude Code through the Anthropic endpoint, and is priced per million tokens as listed on its model page.
| Model | What it is | Context | Price in / out | Supports |
|---|---|---|---|---|
| GLM 5.3 Xploded glm-5.3-xploded | GLM 5.3 uncensored API | 1M | $2.00 / $5.00 | tools, reasoning |
| GLM 5.3 Flash Xploded glm-5.3-flash-xploded | GLM 5.3 Flash uncensored API | 1M | $0.30 / $1.00 | tools, vision, reasoning |
| DeepSeek V4.1 Flash 1017 Xploded deepseek-v4.1-flash-1017-xploded | DeepSeek V4.1 Flash 1017 abliterated API | 1M | $1.25 / $2.50 | tools, vision, reasoning |
| DeepSeek Flash Lite Xploded deepseek-v4.1-flash-lite-1017-xploded | DeepSeek Flash Lite abliterated API | 512K | $0.36 / $1.44 | tools, vision, reasoning |
| MiMo V2.6 Flash Xploded mimo-v2.6-flash-xploded | MiMo V2.6 Flash abliterated API | 1M | $1.00 / $2.00 | tools, vision, reasoning |
| Qwen3.8 27B Xploded qwen3.8-27b-xploded | Qwen3.8 27B uncensored API | 1M | $1.50 / $5.00 | tools, vision, reasoning |
| Qwen3.8 27B Ultrafast Xploded qwen3.8-27b-fast-xploded | Qwen3.8 27B Ultrafast uncensored API | 524K | $1.75 / $5.50 | tools, vision, reasoning |
FAQ
Which uncensored models does MagmaRouter sell?
MagmaRouter sells GLM 5.3 Xploded, GLM 5.3 Flash Xploded, DeepSeek V4.1 Flash 1017 Xploded, DeepSeek Flash Lite Xploded, MiMo V2.6 Flash Xploded, Qwen3.8 27B Xploded, Qwen3.8 27B Ultrafast Xploded. The table above has the price and context for each.
Are these models abliterated or uncensored?
DeepSeek V4.1 Flash 1017 Xploded and DeepSeek Flash Lite Xploded and MiMo V2.6 Flash Xploded are abliterated: the refusal direction was removed from the weights. GLM 5.3 Xploded, GLM 5.3 Flash Xploded, Qwen3.8 27B Xploded, Qwen3.8 27B Ultrafast Xploded are uncensored: the refusal was trained back out. The abliterated vs uncensored page explains the difference.
Which one answers fastest?
MiMo V2.6 Flash Xploded usually starts answering in one to two seconds. The GLM 5.3 Flash model thinks before it answers, so its first word can take much longer.