Phi 3.5 Mini Instruct
Aug 2024
Phi 3.5 Mini InstructCompare inonMMLU-Pro · Official board
—per task?
self-host onlyprice
—output speed?
131Kcontext
Alternatives?
ModelScoreCostPer taskWhat changes
BetterGemini 3.1 Prounstated+90%—$2 / $12OpenGemma 2 9B It+9%—$0.03 / $0.06Context: 131K → 8K+1
BetterGPT-4ounstated+56%—$2.50 / $10Context: 131K → 128KBetterGPT OSS 20BmediumOpen+53%—$0.018 / $0.09Memory (4-bit): 1.9 GB → 13 GB
BetterQwQ 32BOpen+44%—$0.15 / $0.40Memory (4-bit): 1.9 GB → 20 GBModelScoreCostPer taskWhat changes
Nothing here yet.
ModelScoreCostPer taskWhat changes
Gemini 3.1 Prounstated+90%—$2 / $12GPT-4ounstated+56%—$2.50 / $10Context: 131K → 128KGPT OSS 20BmediumOpen+53%—$0.018 / $0.09Memory (4-bit): 1.9 GB → 13 GBQwQ 32BOpen+44%—$0.15 / $0.40Memory (4-bit): 1.9 GB → 20 GB
Llama 3.3 70B InstructunstatedOpen+38%—$0.10 / $0.32Context: 131K → 128K+1ModelScoreCostPer taskWhat changes
Gemma 2 9B ItOpen+9%—$0.03 / $0.06Context: 131K → 8K+1Qwen 2 7B InstructOpensame—$0.055 / $0.055Context: 131K → 33K+1LEXAONE 3.5 7.8B InstructOpen-3%——Context: 131K → 33K+1
Mistral NemounstatedOpen-6%—$0.15 / $0.15Memory (4-bit): 1.9 GB → 6.1 GBLlama 3.1 8B InstructunstatedOpen-8%—$0.02 / $0.04Memory (4-bit): 1.9 GB → 4 GBModelScoreCostPer taskWhat changes
WizardLM 2 8x22Bunstated-18%—$0.48 / $0.48Context: 131K → 66KPhi 4 Reasoning PlusOpenn/a—$0.07 / $0.35Context: 131K → 33K+1Phi 4 ReasoningOpenn/a——Context: 131K → 33K+1Phi 4non-thinkingOpenn/a—$0.056 / $0.11Context: 131K → 16K+1Phi 3 Medium 128K InstructOpenn/a——Memory (4-bit): 1.9 GB → 7 GBPhi 3 Medium 4K InstructOpenn/a—$0.14 / $0.14Context: 131K → 4K+1Phi 3 Small 8K InstructOpenn/a——Context: 131K → 8K+1Phi 3 Mini 128K InstructOpenn/a——All Microsoft models →Run it yourself
estimateFits on✓ 8 GB (phone, laptop) · full quality✓ 16 GB · full quality✓ 24 GB GPU · full quality✓ 48 GB GPU · full quality✓ 80 GB GPU · full quality
LicenceMITfree for commercial use
Benchmarks
in Chat & KnowledgeMMLU-Pro · official board47.9#17
FAQ
How much does Phi 3.5 Mini Instruct cost?›
There is no per-unit API price: self-host only.
Is Phi 3.5 Mini Instruct free to use?›
There is no free API. The weights are open, so you can run it yourself.
How good is Phi 3.5 Mini Instruct, and what is it best at?›
#17 of 32 in Chat & Knowledge.
Is Phi 3.5 Mini Instruct better than Jamba 1.5 Large?›
On MMLU-Pro · official board, Phi 3.5 Mini Instruct scores 47.9 vs 49.5 — lower.
What is the context window of Phi 3.5 Mini Instruct?›
131K tokens.
When was Phi 3.5 Mini Instruct released? Is there a newer version?›
Released 2024-08-16. It is the current version in its line.
Is Phi 3.5 Mini Instruct open source?›
Yes, the weights are open (MIT licence).
Can I run Phi 3.5 Mini Instruct locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Phi 3.5 Mini Instruct read images?›
No, it accepts text only.
What is Phi 3.5 Mini Instruct?›
Phi 3.5 Mini Instruct is an open-weights language model from Microsoft, released Aug 2024.
What are the best alternatives to Phi 3.5 Mini Instruct?›
Gemini 3.1 Pro (better), Gemma 2 9B It (open weights).