Compare · ModelsLive · 3 picked · head to head
Qwen2.5 72B Instruct Abliterated vs Llama 3.3 70B Instruct vs Qwen2.5 72B Instruct
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Qwen2.5 72B Instruct wins on 8/16 benchmarks
Qwen2.5 72B Instruct wins 8 of 16 shared benchmarks. Leads in general · knowledge · math.
Category leads
general·Qwen2.5 72B Instructknowledge·Qwen2.5 72B Instructlanguage·Llama 3.3 70B Instructmath·Qwen2.5 72B Instructreasoning·Llama 3.3 70B Instructcoding·Qwen2.5 72B Instructarena·Llama 3.3 70B Instruct
Hype vs Reality
Attention vs performance
Qwen2.5 72B Instruct Abliterated
#139 by perf·no signal
Llama 3.3 70B Instruct
#218 by perf·#18 by attention
Qwen2.5 72B Instruct
#128 by perf·#2 by attention
Best value
Llama 3.3 70B Instruct
1.3x better value than Qwen2.5 72B Instruct
Qwen2.5 72B Instruct Abliterated
n/a
no price
Llama 3.3 70B Instruct
171.9 pts/$
$0.21/M
Qwen2.5 72B Instruct
129.5 pts/$
$0.38/M
Vendor risk
Who is behind the model
HA
HuiHui AI
private · undisclosed
Meta AI
$1.87T·Tier 1
Alibaba (Qwen)
$293.0B·Tier 1
Head to head
16 benchmarks · 3 models
Qwen2.5 72B Instruct AbliteratedLlama 3.3 70B InstructQwen2.5 72B Instruct
BBH (HuggingFace)
Qwen2.5 72B Instruct leads by +1.4
Qwen2.5 72B Instruct Abliterated
60.5
Llama 3.3 70B Instruct
56.6
Qwen2.5 72B Instruct
61.9
GPQA
Qwen2.5 72B Instruct Abliterated leads by +2.7
Qwen2.5 72B Instruct Abliterated
19.4
Llama 3.3 70B Instruct
10.5
Qwen2.5 72B Instruct
16.7
IFEval
Llama 3.3 70B Instruct leads by +3.6
Qwen2.5 72B Instruct Abliterated
85.9
Llama 3.3 70B Instruct
90.0
Qwen2.5 72B Instruct
86.4
MATH Level 5
Qwen2.5 72B Instruct Abliterated leads by +0.3
Qwen2.5 72B Instruct Abliterated
60.1
Llama 3.3 70B Instruct
48.3
Qwen2.5 72B Instruct
59.8
MMLU-PRO
Qwen2.5 72B Instruct leads by +1.0
Qwen2.5 72B Instruct Abliterated
50.4
Llama 3.3 70B Instruct
48.1
Qwen2.5 72B Instruct
51.4
MUSR
Llama 3.3 70B Instruct leads by +3.2
Qwen2.5 72B Instruct Abliterated
12.3
Llama 3.3 70B Instruct
15.6
Qwen2.5 72B Instruct
11.7
Aider · Code Editing
Qwen2.5 72B Instruct leads by +6.0
Llama 3.3 70B Instruct
59.4
Qwen2.5 72B Instruct
65.4
Chatbot Arena Elo · Overall
Llama 3.3 70B Instruct leads by +15.0
Llama 3.3 70B Instruct
1317.8
Qwen2.5 72B Instruct
1302.8
Balrog
Llama 3.3 70B Instruct leads by +6.8
Balrog · benchmarks AI agents on text-based adventure games, testing language understanding, strategic planning, and long-horizon reasoning.
Llama 3.3 70B Instruct
23.0
Qwen2.5 72B Instruct
16.2
Dtbench
Qwen2.5 72B Instruct leads by +5.8
Llama 3.3 70B Instruct
32.5
Qwen2.5 72B Instruct
38.2
GPQA diamond
Qwen2.5 72B Instruct leads by +2.3
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Llama 3.3 70B Instruct
29.9
Qwen2.5 72B Instruct
32.2
Lmca
Llama 3.3 70B Instruct leads by +4.8
Llama 3.3 70B Instruct
20.6
Qwen2.5 72B Instruct
15.8
MATH level 5
Qwen2.5 72B Instruct leads by +21.6
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
Llama 3.3 70B Instruct
41.6
Qwen2.5 72B Instruct
63.2
MMLU
Llama 3.3 70B Instruct leads by +1.3
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
Llama 3.3 70B Instruct
81.7
Qwen2.5 72B Instruct
80.4
OTIS Mock AIME 2024-2025
Qwen2.5 72B Instruct leads by +2.9
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Llama 3.3 70B Instruct
5.0
Qwen2.5 72B Instruct
8.0
WeirdML
Qwen2.5 72B Instruct leads by +1.5
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Llama 3.3 70B Instruct
14.4
Qwen2.5 72B Instruct
16.0
Full benchmark table
| Benchmark | Qwen2.5 72B Instruct Abliterated | Llama 3.3 70B Instruct | Qwen2.5 72B Instruct |
|---|---|---|---|
BBH (HuggingFace) | 60.5 | 56.6 | 61.9 |
GPQA | 19.4 | 10.5 | 16.7 |
IFEval | 85.9 | 90.0 | 86.4 |
MATH Level 5 | 60.1 | 48.3 | 59.8 |
MMLU-PRO | 50.4 | 48.1 | 51.4 |
MUSR | 12.3 | 15.6 | 11.7 |
Aider · Code Editing | — | 59.4 | 65.4 |
Chatbot Arena Elo · Overall | — | 1317.8 | 1302.8 |
Balrog Balrog · benchmarks AI agents on text-based adventure games, testing language understanding, strategic planning, and long-horizon reasoning. | — | 23.0 | 16.2 |
Dtbench | — | 32.5 | 38.2 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | — | 29.9 | 32.2 |
Lmca | — | 20.6 | 15.8 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | — | 41.6 | 63.2 |
MMLU Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge. | — | 81.7 | 80.4 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | — | 5.0 | 8.0 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | — | 14.4 | 16.0 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
HA Qwen2.5 72B Instruct Abliterated | — | — | — | — |
| $0.10 | $0.32 | 131K tokens (~66 books) | $1.55 | |
| $0.36 | $0.40 | 33K tokens (~16 books) | $3.70 |