LLM

DeepSeek V4 Pro (Reasoning, High Effort)

Name: DeepSeek V4 Pro (Reasoning, High Effort) Review & Benchmarks
Item: DeepSeek V4 Pro (Reasoning, High Effort)
Rating: 43.7
Author: Sentient Weekly

Overall Score

43.7

Released Apr 2026

Benchmark Scores

Reasoning

—

Coding

58.7

Math

—

Creative writing

—

Instruction following

—

Multimodal

—

Standard Benchmarks

aa_terminalbench_v2_1aa_terminalbench_v2_1

0.7

Measured 2026-08-12 · source

aa_gpqaaa_gpqa

0.9

Measured 2026-08-12 · source

aa_hleaa_hle

0.3

Measured 2026-08-12 · source

aa_scicodeaa_scicode

0.5

Measured 2026-08-12 · source

aa_ifbenchaa_ifbench

0.7

Measured 2026-08-12 · source

aa_tau_bankingaa_tau_banking

0.3

Measured 2026-08-12 · source

aa_lcraa_lcr

0.7

Measured 2026-08-12 · source

aa_terminalbench_hardaa_terminalbench_hard

0.4

Measured 2026-08-12 · source

aa_tau2aa_tau2

0.9

Measured 2026-08-12 · source

Price

$1.74/M

Paid

Context

Speed

61.19 t/s

1398ms TTFT

Compare this model →

Discussion

0 comments

Be the first to comment.