vs
0:15
overall capability wins (0 ties across 15 scored). Llama 4 Maverick leads overall.
- Reasoning0:8
- Coding0:3
- Arena0:1
- Pricing2:1
- Speed0:2
- Specs0:3
Verdict
Llama 4 Maverick leads coding (widest gap: +26.5 on LiveCode); Llama 4 Maverick ranks higher on LMArena (+28 Elo); Llama 4 Maverick offers ~7.8x larger context (1M vs 128K).
- Llama 4 Maverick leads coding (widest gap: +26.5 on LiveCode)
- Llama 4 Maverick ranks higher on LMArena (+28 Elo)
- Llama 4 Maverick offers ~7.8x larger context (1M vs 128K)
- Llama 4 Maverick allows ~8.0x more max output (33K vs 4K)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Llama-3.1-Nemotron-70B Instruct
- $0.45
- Llama 4 Maverick
- $0.48
Llama-3.1-Nemotron-70B Instruct is about 1.1x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
NVIDIA
Meta
Family
Llama / Muse
Llama / Muse
License
NVIDIA Open Model
Llama 4 Community
Open weights
Yes
Yes
Release
Oct 15, 2024
Apr 5, 2025
Knowledge cutoff
-
-
API / provider
NVIDIA NIM
Together / Fireworks (ref.)
Modalities
text → text
text, image → text
Specs
Context window
128K
1M
Max output
4K
33K
Parameters
70B
400B (17B act.)
Pricing
Input $/1M
$0.35
$0.27
Output $/1M
$0.40
$0.85
Blended $/1M (3∶1)
$0.36
$0.42
Speed
tok/s
60
95
TTFT (s)
0.45
0.35
Reasoning
MMLU-Pro
69.0
80.5
GPQA Diamond
46.5
67.1
Humanity's Last Exam
4.2
4.9
AIME 2025
11.0
19.3
MATH-500
73.3
88.9
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
17.8
24.9
AA-LCR v1.1
8.3
50.0
CritPt
—
—
MMMU-Pro
—
62.1
IFBench
30.7
43.0
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
21.0
SWE-bench Pro
—
5.2
SWE-bench Multilingual
—
—
LiveCodeBench
16.9
43.4
Terminal-Bench 2.1
—
7.9
Aider Polyglot
—
15.6
Terminal-Bench 3
—
—
BigCodeBench
38.7
49.7
SciCode
23.3
31.7
CursorBench
—
—
SWE-Rebench
—
11.0
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,299
1,327
| Metric | Llama-3.1-Nemotron-70B Instruct | Llama 4 Maverick | Delta |
|---|---|---|---|
| Identity | |||
| Organization | NVIDIA | Meta | — |
| Family | Llama / Muse | Llama / Muse | — |
| License | NVIDIA Open Model | Llama 4 Community | — |
| Open weights | Yes | Yes | — |
| Release | Oct 15, 2024 | Apr 5, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | NVIDIA NIM | Together / Fireworks (ref.) | — |
| Modalities | text → text | text, image → text | — |
| Specs | |||
| Context window | 128K | 1M | B +872K |
| Max output | 4K | 33K | B +29K |
| Parameters | 70B | 400B (17B act.) | B +330 |
| Pricing | |||
| Input $/1M | $0.35 | $0.27 | B +$0.08 |
| Output $/1M | $0.40 | $0.85 | A +$0.45 |
| Blended $/1M (3∶1) | $0.36 | $0.42 | A +$0.05 |
| Speed | |||
| tok/s | 60 | 95 | B +35 |
| TTFT (s) | 0.45 | 0.35 | B +0.10 |
| Reasoning | |||
| MMLU-Pro | 69.0 | 80.5 | B +11.5 pts |
| GPQA Diamond | 46.5 | 67.1 | B +20.6 pts |
| Humanity's Last Exam | 4.2 | 4.9 | B +0.7 pts |
| AIME 2025 | 11.0 | 19.3 | B +8.3 pts |
| MATH-500 | 73.3 | 88.9 | B +15.6 pts |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 17.8 | 24.9 | B +7.1 pts |
| AA-LCR v1.1 | 8.3 | 50.0 | B +41.7 pts |
| CritPt | — | — | — |
| MMMU-Pro | — | 62.1 | — |
| IFBench | 30.7 | 43.0 | B +12.3 pts |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | 21.0 | — |
| SWE-bench Pro | — | 5.2 | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 16.9 | 43.4 | B +26.5 pts |
| Terminal-Bench 2.1 | — | 7.9 | — |
| Aider Polyglot | — | 15.6 | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | 38.7 | 49.7 | B +11.0 pts |
| SciCode | 23.3 | 31.7 | B +8.4 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | 11.0 | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,299 | 1,327 | B +28 Elo |
Llama-3.1-Nemotron-70B Instruct
- Context
- 128K / 4K out
- Parameters
- 70B
- Price
- $0.35 / $0.40
- Speed
- 60 tok/s · 0.45s TTFT
- Modalities
- text → text
- License
- NVIDIA Open Model
NVIDIA's RLHF-tuned Llama 3.1 70B — preference-aligned open instruct model widely served on NIM.
Llama 4 Maverick
- Context
- 1M / 33K out
- Parameters
- 400B (17B act.)
- Price
- $0.27 / $0.85
- Speed
- 95 tok/s · 0.35s TTFT
- Modalities
- text, image → text
- License
- Llama 4 Community
Meta's open multimodal MoE flagship - 400B total / 17B active with 1M context.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.