Phi 4
Phi 4Rankingin non-thinking mode?
Alternatives?
CheaperCommand R7b+11%-6%$0.037 / $0.15Context: not verified
BetterClaude Opus 4.5unstated×2.7+14,186%$5 / $25NOpenNanbeige4 3B Thinking 2511+79%——
CheaperGemma 3 12BunstatedOpen+6%-11%$0.05 / $0.10
CheaperLlama 3.1 8B InstructunstatedOpen-10%-64%$0.02 / $0.04Command R7b+11%-6%$0.037 / $0.15Context: not verifiedGemma 3 12BunstatedOpen+6%-11%$0.05 / $0.10Llama 3.1 8B InstructunstatedOpen-10%-64%$0.02 / $0.04
Qwen3 0.6BunstatedOpen-17%-29%$0.04 / $0.08Llama 3.2 3B InstructunstatedOpen-24%-71%$0.02 / $0.02Claude Opus 4.5unstated×2.7+14,186%$5 / $25Claude Sonnet 4.5unstated×2.5+8,471%$3 / $15Claude Haiku 4.5unstated×2.4+2,757%$1 / $5
o3 (2025-04-16)unstated×2.2+4,900%$2 / $8
Kimi K2 0711unstatedOpen×2.1+1,150%$0.50 / $2Memory (4-bit): 7.3 GB → 616 GB+1Qwen3 32BunstatedOpen+69%+1,650%$0.70 / $2.80Memory (4-bit): 7.3 GB → 20 GBQwen3 8BunstatedOpen+48%+343%$0.18 / $0.70Qwen3 30B A3B Instruct 2507unstatedOpen+44%+104%$0.09 / $0.30Memory (4-bit): 7.3 GB → 18 GBQwen3 14BunstatedOpen+43%+775%$0.35 / $1.40Memory (4-bit): 7.3 GB → 7.4 GBPhi 4 Reasoning PlusOpenn/a—$0.07 / $0.35Phi 4 ReasoningOpenn/a——Phi 3.5 Mini InstructOpenn/a——Phi 3 Medium 128K InstructOpenn/a——Phi 3 Medium 4K InstructOpenn/a—$0.14 / $0.14Context: 16K → 4KPhi 3 Small 8K InstructOpenn/a——Context: 16K → 8KPhi 3 Mini 128K InstructOpenn/a——Phi 3 Mini 4K InstructOpenn/a—$0.40 / $0.80Context: 16K → 4KAll Microsoft models →Run it yourself
estimateBenchmarks
all categoriesFAQ
How much does Phi 4 cost?›
$0.056 per 1M input tokens and $0.11 per 1M output tokens (cheapest host: DeepInfra).
Is Phi 4 free to use?›
There is no free API; the cheapest paid price is $0.056 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Phi 4, and what is it best at?›
#316 of 334 in Chat & Knowledge, #171 of 239 in Coding, #229 of 242 in Agents & Tool Use, #171 of 195 in Writing & Language.
Is Phi 4 better than Gemma 3 27B?›
On BFCL v4, Phi 4 scores 28.8 vs 29.5 — about the same.
How can I use Phi 4?›
Through the API of Azure AI Foundry, DeepInfra, Featherless.
What is the context window of Phi 4?›
16K tokens, with up to 15K output tokens.
When was Phi 4 released? Is there a newer version?›
Released 2024-12-12. It is the current version in its line.
Is Phi 4 open source?›
Yes, the weights are open (MIT licence).
Can I run Phi 4 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Phi 4 read images?›
No, it accepts text only.
What is Phi 4?›
Phi 4 is an open-weights language model from Microsoft, released Dec 2024.
What are the best alternatives to Phi 4?›
Command R7b (cheaper), Claude Opus 4.5 (better), Nanbeige4 3B Thinking 2511 (open weights).