VerdictPal · editorial desk · 2026VerdictPal

Compare · Two models, head-to-head

Pick two models. See the diff.

Composite scores, benchmark ranks, capability grids, and pricing per million tokens. All sourced from the same data that powers the model atlas.

Cards Models

DeepSeek V4 Pro Max vs Qwen 3.7 Max

Model A

Model B

MODEL A

DeepSeek V4 Pro Max

DeepSeek · DeepSeek

Composite 74Intel 5614 rows10 sources

DeepSeek's April 2026 frontier open-weights model, leading the open-weights tier on SWE-bench Verified (80.6%) and Terminal-Bench 2.0 (67.9%) at $0.40 / $1.60 per 1M tokens. Cheapest frontier pricing on the public market.

MODEL B

Qwen 3.7 Max

Alibaba · Qwen

Composite 73Intel 4814 rows9 sources

Alibaba's May 2026 flagship open-weights model, matching frontier reasoning at open-weight price points. AA Intelligence Index 56.6 (#6 globally), Terminal-Bench 2.0 69.7%. 128K context, $0.40 / $1.20 per 1M tokens.

DeepSeek V4 Pro Max4stats won

Tied1stats

Qwen 3.7 Max0stats won

Spec showdown

Capacity, price, freshness.

DeepSeek V4 Pro MaxvsQwen 3.7 Max

Composite scores are within 5 points — 74 for DeepSeek V4 Pro Max vs 73 for Qwen 3.7 Max. The choice comes down to capability profile and price. DeepSeek V4 Pro Max has the higher Vals/AA intelligence score (56 vs 48).

Composite score
DeepSeek V4 Pro Max74Winner on this rowQwen 3.7 Max73
Intelligence (Vals/AA)
DeepSeek V4 Pro Max56Winner on this rowQwen 3.7 Max48
Context window
DeepSeek V4 Pro Max256KWinner on this rowQwen 3.7 Max128K
Max output
DeepSeek V4 Pro Max66KWinner on this rowQwen 3.7 Max33K
Input price / 1M (lower wins)
DeepSeek V4 Pro Max$0.40Qwen 3.7 Max$0.40

Composite score

Intelligence (Vals/AA)

Context window

256K

128K

Max output

66K

33K

Input price / 1M (lower wins)

$0.40

Capability grid

What each model can do.

Modalities

DeepSeek V4 Pro MaxQwen 3.7 Max

Text●●

Image——

Audio——

Video——

Code●●

Tool use●●

Reasoning●●

Access

DeepSeek V4 Pro MaxQwen 3.7 Max

API●●

Consumer app——

Open weights●●

Managed cloud●●

On-device——

Family radar

Top score per benchmark family.

Knowledge

Reasoning

Math

Coding

Agentic

Long context

Tool use

Human preference

Benchmark ranks

Every shared benchmark, side by side.

Knowledge

MMLU#8/22 · #10/22

89.7%

88.9%

MMLU-Pro#14/21 · #16/21

76.4%

75.1%

Reasoning

GPQA Diamond#56/69 · #60/69

75.6%

73.8%

LiveBench#14/21 · #15/21

61.9%

61.6%

Math

MATH#12/20 · #15/20

83.9%

81.6%

Coding

HumanEval#9/15 · #10/15

93.4% pass@1

93.1% pass@1

SWE-bench Verified#7/23 · #14/23

80.6%

79.4%

Agentic

Terminal-Bench#25/69 · #22/69

67.9%

69.7%

Vals Index#9/20 · #14/20

56.23%

48.04%

Long context

LongBench v2#14/19 · #16/19

48.6%

47.8%

Tool use

BFCL#16/20 · #10/20

74.6%

78.4%

MCP Atlas#6/10 · #7/10

72.6%

71.8%

Human preference

LMArena (Chatbot Arena)#5/13 · #7/13

1418

1409

Dossier facts

Dates, sources, fineprint.

DeepSeek V4 Pro MaxQwen 3.7 Max

ProviderDeepSeekAlibaba

FamilyDeepSeekQwen

Release date2026-04-242026-05-20

Knowledge cutoff2026-02-012026-02-01

Context window256K128K

Max output66K33K

Input price /1M$0.40$0.40

Output price /1M$1.6$1.2

Statusflagshipflagship

Composite score7473

Intelligence score5648

Intel sourcevalsvals

Benchmark rows1414

Source count109

Freshnessfreshfresh

Pricing checked2026-06-012026-06-01

Ecosystem

Which tools wrap each model.

DeepSeek V4 Pro Max

deepseek

Qwen 3.7 Max

Not listed in any tool yet.

Verdict

Which model to actually pick.

Open DeepSeek V4 Pro Max's dossier Open Qwen 3.7 Max's dossier