Compare · ModelsLive · 2 picked · head to head
phi-3-medium 14B vs GPT-3.5 Turbo (older v0613)
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
phi-3-medium 14B wins on 6/9 benchmarks
phi-3-medium 14B wins 6 of 9 shared benchmarks. Leads in knowledge · reasoning · math.
Category leads
knowledge·phi-3-medium 14Breasoning·phi-3-medium 14Bmath·phi-3-medium 14B
Hype vs Reality
Attention vs performance
phi-3-medium 14B
#48 by perf·no signal
GPT-3.5 Turbo (older v0613)
#111 by perf·no signal
Best value
GPT-3.5 Turbo (older v0613)
phi-3-medium 14B
—
no price
GPT-3.5 Turbo (older v0613)
30.5 pts/$
$1.50/M
Vendor risk
Who is behind the model
Microsoft
$3.00T·Big Tech
OpenAI
$840.0B·Tier 1
Head to head
9 benchmarks · 2 models
phi-3-medium 14BGPT-3.5 Turbo (older v0613)
ANLI
GPT-3.5 Turbo (older v0613) leads by +3.4
ANLI (Adversarial NLI) · adversarially constructed natural language inference dataset where each round targets weaknesses found in previous model generations.
phi-3-medium 14B
33.7
GPT-3.5 Turbo (older v0613)
37.1
ARC AI2
phi-3-medium 14B leads by +5.6
AI2 Reasoning Challenge · tests grade-school level science knowledge with multiple-choice questions requiring reasoning beyond simple retrieval.
phi-3-medium 14B
88.8
GPT-3.5 Turbo (older v0613)
83.2
BBH
phi-3-medium 14B leads by +26.4
BIG-Bench Hard · a curated subset of 23 challenging tasks from BIG-Bench where language models previously failed to outperform average humans.
phi-3-medium 14B
75.2
GPT-3.5 Turbo (older v0613)
48.8
GPQA diamond
phi-3-medium 14B leads by +0.6
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
phi-3-medium 14B
3.5
GPT-3.5 Turbo (older v0613)
2.9
MATH level 5
phi-3-medium 14B leads by +5.9
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
phi-3-medium 14B
17.6
GPT-3.5 Turbo (older v0613)
11.6
MMLU
phi-3-medium 14B leads by +14.3
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
phi-3-medium 14B
70.7
GPT-3.5 Turbo (older v0613)
56.4
OpenBookQA
phi-3-medium 14B leads by +1.9
OpenBookQA · science questions that require combining a given core fact with broad common knowledge, mimicking an open-book exam setting.
phi-3-medium 14B
83.2
GPT-3.5 Turbo (older v0613)
81.3
TriviaQA
GPT-3.5 Turbo (older v0613) leads by +11.9
TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents.
phi-3-medium 14B
73.9
GPT-3.5 Turbo (older v0613)
85.8
Winogrande
GPT-3.5 Turbo (older v0613) leads by +0.2
WinoGrande · large-scale commonsense reasoning benchmark where models must resolve ambiguous pronouns in carefully constructed sentence pairs.
phi-3-medium 14B
63.0
GPT-3.5 Turbo (older v0613)
63.2
Full benchmark table
| Benchmark | phi-3-medium 14B | GPT-3.5 Turbo (older v0613) |
|---|---|---|
ANLI ANLI (Adversarial NLI) · adversarially constructed natural language inference dataset where each round targets weaknesses found in previous model generations. | 33.7 | 37.1 |
ARC AI2 AI2 Reasoning Challenge · tests grade-school level science knowledge with multiple-choice questions requiring reasoning beyond simple retrieval. | 88.8 | 83.2 |
BBH BIG-Bench Hard · a curated subset of 23 challenging tasks from BIG-Bench where language models previously failed to outperform average humans. | 75.2 | 48.8 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 3.5 | 2.9 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 17.6 | 11.6 |
MMLU Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge. | 70.7 | 56.4 |
OpenBookQA OpenBookQA · science questions that require combining a given core fact with broad common knowledge, mimicking an open-book exam setting. | 83.2 | 81.3 |
TriviaQA TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents. | 73.9 | 85.8 |
Winogrande WinoGrande · large-scale commonsense reasoning benchmark where models must resolve ambiguous pronouns in carefully constructed sentence pairs. | 63.0 | 63.2 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| — | — | — | — | |
| $1.00 | $2.00 | 4K tokens (~2 books) | $12.50 |