Phi 3 Medium 4K Instruct
May 2024
Phi 3 Medium 4K InstructCompare inon
Benchmarks of Phi 3 Medium 4K Instruct16
BenchmarkByScoreRank
No benchmark matches
Ranking?
Alternatives?
ModelScoreCostPer taskWhat changes
CheaperQwen 3.5 Flashunstatedwins 79%-19%$0.065 / $0.26
BetterGPT-5 Minihighwins 77%+391%$0.053
OpenGemma 4 31Bunstatedwins 84%+14%$0.10 / $0.34Memory (4-bit): 7 GB → 19 GB
CheaperNemotron 3 Nano 30B A3BunstatedOpenwins 77%-25%$0.06 / $0.24Memory (4-bit): 7 GB → 19 GBCheaperGemma 3 27BunstatedOpenwins 77%-29%$0.08 / $0.16Memory (4-bit): 7 GB → 16 GBModelScoreCostPer taskWhat changes
Qwen 3.5 Flashunstatedwins 79%-19%$0.065 / $0.26Nemotron 3 Nano 30B A3BunstatedOpenwins 77%-25%$0.06 / $0.24Memory (4-bit): 7 GB → 19 GBGemma 3 27BunstatedOpenwins 77%-29%$0.08 / $0.16Memory (4-bit): 7 GB → 16 GBGPT OSS 120BunstatedOpenwins 77%-57%$0.03 / $0.15Memory (4-bit): 7 GB → 70 GBGemma 3 12BunstatedOpenwins 74%-55%$0.05 / $0.10ModelScoreCostPer taskWhat changes
GPT-5 Minihighwins 77%+391%$0.053GPT-6 Lunamaxwins 79%+43%$0.068Function calling only with none · Chat Completions+3
Claude Haiku 5.5highwins 80%+43%$0.079Remove sampling parameters+5GPT-5.6 Lunaxhighwins 83%+221%$0.085GPT-5.6 Terraxhighwins 83%+3,114%$0.63Prompts with >272K input tokensModelScoreCostPer taskWhat changes
Gemma 4 31BunstatedOpenwins 84%+14%$0.10 / $0.34Memory (4-bit): 7 GB → 19 GBQwen3.8 27BunstatedOpenwins 83%+583%$0.42 / $2.55Memory (4-bit): 7 GB → 17 GB+5Gemma 4 26B A4BunstatedOpenwins 83%+87%$0.15 / $0.60Memory (4-bit): 7 GB → 15 GBQwen 3.5 27BunstatedOpenwins 81%+489%$0.30 / $2.40Memory (4-bit): 7 GB → 17 GBQwen3 30B A3B Instruct 2507unstatedOpenwins 78%same$0.09 / $0.30Memory (4-bit): 7 GB → 18 GBModelScoreCostPer taskWhat changes
Phi 4unstatedOpenwins 59%—$0.056 / $0.11Memory (4-bit): 7 GB → 7.3 GBPhi 3 Small 8K InstructOpenwins 47%——Phi 3 Mini 4K InstructOpenwins 43%—$0.40 / $0.80Phi 3 Mini 128K InstructOpenwins 38%——Phi 4 Reasoning PlusOpenn/a—$0.07 / $0.35Memory (4-bit): 7 GB → 7.3 GBPhi 4 ReasoningOpenn/a——Memory (4-bit): 7 GB → 7.3 GBPhi 3.5 Mini InstructOpenn/a——Phi 3 Medium 128K InstructOpenn/a——All Microsoft models →Run it yourself
estimateFits on✓ 8 GB (phone, laptop) · compressed✓ 16 GB · 8-bit✓ 24 GB GPU · 8-bit✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
LicenceMITfree for commercial use
Benchmarks
all categoriesChat & Knowledge
Arena Text · overall1,198#183
Arena Text · Chinese1,108.2#187
Arena Text · English1,169.2#194
Arena Text · French1,110.6#158
Arena Text · German1,100.8#158
Arena Text · instruction following1,113.3#193
Arena Text · Japanese1,042.4#142
Arena Text · Korean953.9#153
Arena Text · non-English1,093.8#193
Arena Text · Russian1,144.8#190
Arena Text · Spanish1,095.1#164
Arena Text · overall · October 22, 20261,137.5#186
Arena Text · overall · October 15, 20261,137.5#176
Writing & Language
Arena Text · creative writing1,150#184
FAQ
How much does Phi 3 Medium 4K Instruct cost?›
$0.14 per 1M input tokens and $0.14 per 1M output tokens (cheapest host: DeepInfra).
Is Phi 3 Medium 4K Instruct free to use?›
There is no free API; the cheapest paid price is $0.14 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Phi 3 Medium 4K Instruct, and what is it best at?›
#194 of 205 in Chat & Knowledge, #187 of 197 in Writing & Language.
Is Phi 3 Medium 4K Instruct better than Command R?›
On Arena Text · English, Phi 3 Medium 4K Instruct scores 1,169.2 vs 1,174.2 — about the same.
How can I use Phi 3 Medium 4K Instruct?›
Through the API of DeepInfra.
What is the context window of Phi 3 Medium 4K Instruct?›
4K tokens.
When was Phi 3 Medium 4K Instruct released? Is there a newer version?›
Released 2024-05-07. It is the current version in its line.
Is Phi 3 Medium 4K Instruct open source?›
Yes, the weights are open (MIT licence).
Can I run Phi 3 Medium 4K Instruct locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Phi 3 Medium 4K Instruct read images?›
No, it accepts text only.
What is Phi 3 Medium 4K Instruct?›
Phi 3 Medium 4K Instruct is an open-weights language model from Microsoft, released May 2024.
What are the best alternatives to Phi 3 Medium 4K Instruct?›
Qwen 3.5 Flash (cheaper), GPT-5 Mini (better), Gemma 4 31B (open weights).