MagmaRouter vs DeepInfra
DeepInfra is a large inference host that runs its own GPUs, with a big catalogue, low per-token prices and both OpenAI and Anthropic compatible endpoints. MagmaRouter is a much smaller, focused uncensored line that can drive coding agents, with spend caps you set per key.
Facts checked October 2026 against DeepInfra’s own site, docs and policies.
| MagmaRouter | DeepInfra | |
|---|---|---|
| What it is | Hosted gateway for one uncensored, tool capable model line (Xploded). | An inference platform that serves open and proprietary models through an API on its own hardware, and rents dedicated GPUs for custom deployments. |
| Pricing | Pay as you go, $5 minimum top-up, no subscription required. | Per token for language models (for example DeepSeek-V4-Flash at $0.06 input and $0.18 output per 1M tokens, Llama-3.3-70B-Instruct-Turbo at $0.10 and $0.32), execution time for some other models, and hourly GPUs for custom deployments (A100 80GB $0.89 per hour, H100 80GB $2.20 per hour). No contracts or upfront costs, but a card or prepayment is required. No free tier mentioned; minimum top-up not documented. |
| OpenAI-compatible | Yes | Yes, an OpenAI-compatible chat completions API for all LLM models at https://api.deepinfra.com/v1/openai, with tools and tool_choice supported. |
| Anthropic-compatible | Yes, /v1/messages (verified with Claude Code) | Yes, an Anthropic-compatible Messages API at https://api.deepinfra.com/anthropic (POST /v1/messages and /v1/messages/count_tokens), documented for Claude Code. They note not every Anthropic-specific feature may be supported. |
| Uncensored / abliterated | Five uncensored models, the Xploded line: GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash, MiMo V2.6 Flash and Qwen3.8 27B. Every one calls tools. | No models are labelled uncensored or abliterated, but the catalogue includes roleplay finetunes such as Sao10K L3.1 70B Euryale v2.2, described as a model focused on creative roleplay. The terms ban illegal, fraudulent or deceptive use but list no specific content categories such as sexual material. |
| Prompt retention | No prompt or completion stored; model, tokens and cost kept 90 days; signed prompt-hash receipt | Docs say input is not stored to disk during inference, exists only in memory and is deleted once inference completes; only metadata is logged, and API data is not used for training. Exceptions: bulk inference may be stored encrypted for a short period, and Google and Anthropic models follow those companies' own retention. |
| Per-key spend caps | Lifetime and monthly per-key caps, enforced by reserving the worst case and clamping max_tokens | 200 concurrent requests per model by default. An account spending limit in payment settings, and scoped JWT tokens can be limited by model, expiry (up to 1 year) and maximum USD spend. Ordinary per-API-key budgets are not documented. |
Where DeepInfra is the better choice
- Far larger catalogue at very low per-token prices, including mainstream open models, proprietary models and roleplay finetunes like Euryale, plus hourly dedicated GPUs (H100 from $2.20 per hour) for running your own model.
- They run their own inference hardware and document a clear no-storage, no-training policy for standard models, so you deal with the host directly.
Where MagmaRouter is different
- MagmaRouter sells only a short line of uncensored Xploded models (GLM 5.3 Xploded, GLM 5.3 Flash Xploded, DeepSeek V4.1 Flash Xploded, MiMo V2.6 Flash Xploded, Qwen3.8 27B Xploded), every one able to do tool calling for agents like Claude Code, Cline and OpenCode.
- MagmaRouter puts hard lifetime and monthly spend caps on each key, enforced by reserving the worst-case cost and clamping max_tokens before the call, with a $5 minimum top-up and no subscription.
The uncensored models MagmaRouter sells
If you are comparing DeepInfra 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 uncensored, MiMo V2.6 Flash abliterated, Qwen3.8 27B uncensored and Qwen3.8 27B Fast 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 Xploded deepseek-v4.1-flash-xploded | DeepSeek V4.1 Flash uncensored API | 1M | $0.30 / $1.50 | tools, 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 Fast Xploded qwen3.8-27b-fast-xploded | Qwen3.8 27B Fast uncensored API | 524K | $1.50 / $5.00 | tools, vision, reasoning |
FAQ
Is DeepInfra cheaper than MagmaRouter?
Usually yes on a per-token basis. DeepInfra runs its own GPUs and lists models like DeepSeek-V4-Flash at $0.06 input per 1M tokens. MagmaRouter is priced for its uncensored, tool-calling line and per-key caps, not for the lowest per-token price, so for stock models DeepInfra is cheaper.
Can I use Claude Code with DeepInfra?
Yes. DeepInfra documents an Anthropic-compatible endpoint at https://api.deepinfra.com/anthropic that you set as ANTHROPIC_BASE_URL. MagmaRouter also has an Anthropic-compatible /v1/messages endpoint, so both work with Claude Code.
Does DeepInfra have uncensored models?
It does not label any model uncensored or abliterated, but it hosts roleplay finetunes such as L3.1 70B Euryale v2.2. MagmaRouter's whole line is uncensored and built to handle tool calling for coding agents.
Which uncensored models does MagmaRouter sell?
MagmaRouter sells GLM 5.3 Xploded, GLM 5.3 Flash Xploded, DeepSeek V4.1 Flash Xploded, MiMo V2.6 Flash Xploded, Qwen3.8 27B Xploded, Qwen3.8 27B Fast Xploded. The table above has the price and context for each.
Are these models abliterated or uncensored?
MiMo V2.6 Flash Xploded is abliterated: the refusal direction was removed from the weights. GLM 5.3 Xploded, GLM 5.3 Flash Xploded, DeepSeek V4.1 Flash Xploded, Qwen3.8 27B Xploded, Qwen3.8 27B Fast 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.