DeepInfra
Hacker ApprovedRanked 2 of 14 inference providers by recommenders
Recommended for cheap, fast access to many open-source models through an OpenAI-compatible API. Critics say it quantizes some models and sometimes runs them worse.
Recommended for cheap, fast access to many open-source models through an OpenAI-compatible API. Critics say it quantizes some models and sometimes runs them worse.
47 people gave an opinion
- recommend it
- 36
- criticize it
- 11
- approval share
- 77%
Recommended in 11 of 13 months
- November 2025: not recommended
- December 2025: recommended
- January 2026: recommended
- February 2026: recommended
- March 2026: recommended
- April 2026: recommended
- May 2026: recommended
- June 2026: recommended
- July 2026: recommended
- August 2026: recommended
- September 2026: recommended
- October 2026: recommended
- November 2026: not recommended (month in progress)
Why it carries the badge
- 36 people recommend itNeeds 10 or more, met.
- 3.3 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 11 of 13 monthsNeeds 4 or more, met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Cheap prices for open models
“offer a wide range of models like gpt oss that are very capable and far cheaper than the models from big 3.”
Comment from 7 Jun 2026 in “The OnlyFans Economy of American AI” “I use GLM 4.7 with DeepInfra.com and it's extremely reasonable, though maybe a bit on the slower side.”
Comment from 20 Jan 2026 in “GLM-4.7-Flash” “run the same open models for far cheaper and faster than you could ever build out at home”
Comment from 22 Dec 2025 on Hacker News - Comment from 14 Sep 2026 in “Big AI sets out its terms for regulatory capture”
- Comment from 7 Jul 2026 in “AI: The ROI Runway Could Be Long Outside the Tech Sector”
- Comment from 21 Sep 2026 in “MiMo v2.6”
- Comment from 12 Mar 2026 in “IonRouter (YC W26) – High throughput, low cost inference”
- Comment from 20 Aug 2026 in “Clean up Claude 5's token vomit with a separate LLM”
- Comment from 21 Sep 2026 in “MiMo v2.6”
Wide range of open models
“offer a wide range of models like gpt oss that are very capable and far cheaper than the models from big 3.”
Comment from 7 Jun 2026 in “The OnlyFans Economy of American AI” “try Fireworks AI or DeepInfra who do support more recent models”
Comment from 3 Aug 2026 in “Qwen3.8-Max: A New Bar for Coding and Cowork” “because they host open-source models that are good.”
Comment from 2 Apr 2026 in “Qwen3.6-Plus: Towards Real World Agents” - Comment from 15 Jun 2026 in “Ask HN: Show your AI coded games [June 2026]”
- Comment from 2 Apr 2026 in “Qwen3.6-Plus: Towards Real World Agents”
Fast throughput on most models
“I'm very happy with Deepinfra. Less model coverage, but good prices and quite fast.”
Comment from 14 Sep 2026 in “Big AI sets out its terms for regulatory capture” “the prices are decent for US-based deployments and there's acceptable throughput”
Comment from 7 Jul 2026 in “AI: The ROI Runway Could Be Long Outside the Tech Sector” “I pretty reliably get 200 tokens per second on deep infra.”
Comment from 26 Oct 2026 in “OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?” US-based with good terms
“the prices are decent for US-based deployments and there's acceptable throughput”
Comment from 7 Jul 2026 in “AI: The ROI Runway Could Be Long Outside the Tech Sector” “Their whole catalogue is openai compat. US based.”
Comment from 31 Jul 2026 in “DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis” “DeepInfra has excellent terms of service too!”
Comment from 28 Aug 2026 in “GLM-5.3 is now open-weight” - Comment from 18 Aug 2026 in “GPT-5.6 Sol Pricing Cut by 50%”
What people warn about
Each mark is one comment
Quantized models, possibly lower quality
“Some of them more heavily quantize their models (deepinfra) so take that into your workload considerations.”
Comment from 17 Oct 2026 in “DeepSeek-v4.1-Flash-1017” “it's at FP4 quant. How much that effects the quality of inference for GLM, I can't say.”
Comment from 21 Jul 2026 in “Kimi K3, Qwen 3.8, and Anthropic's (Potential) Unravelling” “has a bad history of quantizing models”
Comment from 9 Oct 2026 in “Why isn't the industry freaking out about DeepSeek 4.1 Flash?” - Comment from 27 Oct 2026 in “Root shell on Samsung TV”
Does not preserve model thinking
“Deepinfra's implementation of it is not correct. Thinking is not preserved, and they're not responding to my submitted issue about it.”
Comment from 20 Apr 2026 in “Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving” “Deepinfra for example is not preserving thinking correctly for GLM5.1”
Comment from 20 Apr 2026 in “Kimi K2.6: Advancing Open-Source Coding” Slower than some providers
“I use GLM 4.7 with DeepInfra.com and it's extremely reasonable, though maybe a bit on the slower side.”
Comment from 20 Jan 2026 in “GLM-4.7-Flash” “deepinfra.com is cheaper for that but its noticeably slower”
Comment from 23 Oct 2026 in “Kimi K3.1 scores 55 on Artificial Analysis, placed on the Pareto line” Weak cache hit rate and pricing
“their cache hit rate is 67%. In comparison the provider with the highest hit rate is at 95%.”
Comment from 28 Aug 2026 in “GLM-5.3 is now open-weight” “DeepInfra's Cache Read is 6.5 times more expensive than DeepSeek.”
Comment from 3 Aug 2026 in “Qwen3.8-Max: A New Bar for Coding and Cowork”
Other inference providers people recommend
The four most recommended other inference providers, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| Fireworks AI | 67 | 19 | 78% | 11 | — | Approved |
| DeepInfraThis page | 36 | 11 | 77% | 11 | — | Approved |
| Together AI | 20 | 7 | 74% | 9 | — | Approved |
| fal.ai | 19 | 7 | 73% | 8 | — | Approved |
| Neuralwatt | 18 | 1 | 95% | 7 | — | Approved |
What people said
40 of the 64 comments that mention it, newest first. Each links to the original comment.
“The same model via Deepinfra and open router (despite setting routing to "exact") was far worse.”
“I pretty reliably get 200 tokens per second on deep infra.”
“deepinfra.com is cheaper for that but its noticeably slower”
“DeepSeek's fork of that will surely improve the self hosting landscape”
“Deepseek just gave every third-party inference provider a more efficient stack for free.”
“a reference implementation of some of DeepSeek's own hard-won inference serving techniques”
“Less efficiently, I'm sure, but on much broader hardware”
“I feel the KV cache system is basically a must for any serious inference hosting”
“has huge implications in the global AI hardware market”
“If that's the thing that makes $0.00X cached inputs possible, then this is huge!”
“It's releasing open source code. This is a good thing for the world”
“Based on this framework, we have achieved high-throughput, low-latency large-scale inference services”
“Some of them more heavily quantize their models (deepinfra) so take that into your workload considerations.”
“has a bad history of quantizing models”
“It is one of my three pinned providers for deepseek v4 flash along with Parasail and Deepinfra.”
“simply because deepinfra.com is so cheap”
“It's pretty amazing what it can do for what it costs at deepinfra.com”
“I'm very happy with Deepinfra. Less model coverage, but good prices and quite fast.”
“DeepInfra has excellent terms of service too!”
“their cache hit rate is 67%. In comparison the provider with the highest hit rate is at 95%.”
“being the first 3rd party provider for it on OpenRouter”
“discounted on flex pricing on deepinfra.com”
“Deepinfra is US based and has ZDR FYI.”
“try Fireworks AI or DeepInfra who do support more recent models”
“DeepInfra's Cache Read is 6.5 times more expensive than DeepSeek.”
“Their whole catalogue is openai compat. US based.”
“it's at FP4 quant. How much that effects the quality of inference for GLM, I can't say.”
“the prices are decent for US-based deployments and there's acceptable throughput”
“models from Kimi/Gemma/DeepSeek, via DeepInfra”
“offer a wide range of models like gpt oss that are very capable and far cheaper than the models from big 3.”
“Run Deepseek on Deepinfra then? Or Fireworks if US-based is important.”
“there is special mark for Data Retention for each provider”
“Deepinfra for example is not preserving thinking correctly for GLM5.1”
“Deepinfra's implementation of it is not correct. Thinking is not preserved, and they're not responding to my submitted issue about it.”
“they host open-source models that are good”
“because they host open-source models that are good.”
“MiniMax where DeepInfra and Together are much cheaper”
“This is a model you use via a cheap API provider like DeepInfra”
“I use GLM 4.7 with DeepInfra.com and it's extremely reasonable, though maybe a bit on the slower side.”
“run the same open models for far cheaper and faster than you could ever build out at home”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.