Compare · ModelsLive · 2 picked · head to head
Qwen2.5 72B Instruct Abliterated vs Phi 4
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Qwen2.5 72B Instruct Abliterated wins on 6/6 benchmarks
Qwen2.5 72B Instruct Abliterated wins 6 of 6 shared benchmarks. Leads in general · knowledge · language.
Category leads
general·Qwen2.5 72B Instruct Abliteratedknowledge·Qwen2.5 72B Instruct Abliteratedlanguage·Qwen2.5 72B Instruct Abliteratedmath·Qwen2.5 72B Instruct Abliteratedreasoning·Qwen2.5 72B Instruct Abliterated
Hype vs Reality
Attention vs performance
Qwen2.5 72B Instruct Abliterated
#139 by perf·no signal
Phi 4
#196 by perf·#20 by attention
Vendor risk
Who is behind the model
HA
HuiHui AI
private · undisclosed
Microsoft
$3.84T·Big Tech
Head to head
6 benchmarks · 2 models
Qwen2.5 72B Instruct AbliteratedPhi 4
BBH (HuggingFace)
Qwen2.5 72B Instruct Abliterated leads by +4.8
Qwen2.5 72B Instruct Abliterated
60.5
Phi 4
55.7
GPQA
Qwen2.5 72B Instruct Abliterated leads by +10.2
Qwen2.5 72B Instruct Abliterated
19.4
Phi 4
9.2
IFEval
Qwen2.5 72B Instruct Abliterated leads by +21.1
Qwen2.5 72B Instruct Abliterated
85.9
Phi 4
64.8
MATH Level 5
Qwen2.5 72B Instruct Abliterated leads by +14.9
Qwen2.5 72B Instruct Abliterated
60.1
Phi 4
45.2
MMLU-PRO
Qwen2.5 72B Instruct Abliterated leads by +2.1
Qwen2.5 72B Instruct Abliterated
50.4
Phi 4
48.3
MUSR
Qwen2.5 72B Instruct Abliterated leads by +0.9
Qwen2.5 72B Instruct Abliterated
12.3
Phi 4
11.4
Full benchmark table
| Benchmark | Qwen2.5 72B Instruct Abliterated | Phi 4 |
|---|---|---|
BBH (HuggingFace) | 60.5 | 55.7 |
GPQA | 19.4 | 9.2 |
IFEval | 85.9 | 64.8 |
MATH Level 5 | 60.1 | 45.2 |
MMLU-PRO | 50.4 | 48.3 |
MUSR | 12.3 | 11.4 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
HA Qwen2.5 72B Instruct Abliterated | — | — | — | — |
| $0.07 | $0.14 | 16K tokens (~8 books) | $0.88 |