Claude Opus 4.6$11.000/MClaude Opus 4.5$33.000/MClaude Sonnet 3.7$6.600/MClaude Opus 3$33.000/MClaude 2.1$12.800/MClaude 2$12.800/MGPT-5$3.875/MGPT-4.5$97.500/MGPT-4 Turbo Preview$16.000/MGPT-4$39.000/MGPT-4-32k$78.000/Mo3$19.000/Mo3-mini$2.090/Mo4-mini$2.090/Mo1$28.500/Mo1-mini$5.700/Mo1-preview$28.500/MGemini 2.5 Pro$3.875/MGemini 1.5 Pro$2.375/MGemini 1.0 Ultra$12.000/MGemini 1.0 Pro$0.800/MPaLM 2 Bison$0.500/MPaLM 2 Unicorn$5.000/MGemma 3 27B$0.270/MGrok 3$6.600/MGrok 2$4.400/MGrok 1.5$8.000/MDeepSeek-V3$0.519/MDeepSeek-V3-0324$0.519/MDeepSeek-R1$1.042/MClaude Opus 4.6$11.000/MClaude Opus 4.5$33.000/MClaude Sonnet 3.7$6.600/MClaude Opus 3$33.000/MClaude 2.1$12.800/MClaude 2$12.800/MGPT-5$3.875/MGPT-4.5$97.500/MGPT-4 Turbo Preview$16.000/MGPT-4$39.000/MGPT-4-32k$78.000/Mo3$19.000/Mo3-mini$2.090/Mo4-mini$2.090/Mo1$28.500/Mo1-mini$5.700/Mo1-preview$28.500/MGemini 2.5 Pro$3.875/MGemini 1.5 Pro$2.375/MGemini 1.0 Ultra$12.000/MGemini 1.0 Pro$0.800/MPaLM 2 Bison$0.500/MPaLM 2 Unicorn$5.000/MGemma 3 27B$0.270/MGrok 3$6.600/MGrok 2$4.400/MGrok 1.5$8.000/MDeepSeek-V3$0.519/MDeepSeek-V3-0324$0.519/MDeepSeek-R1$1.042/M

DeepSeekEfficientLIVE INDEX

DeepSeek V3.2

Name: DeepSeek V3.2
Brand: DeepSeek
Price: 0.252000 USD

text->text

Open-weights frontier model that matched GPT-4-class quality at a fraction of the inference cost. Sparked the late-2024 cost reset.

Profile inherited from upstream DeepSeek-V3 ↗ — this is a hosted variant of the same open-weights model.

Modalities textOfficial model page ↗Provider pricing ↗API docs ↗Pricing source ↗Compare with another model →Estimate monthly cost →

INPUT

$0.252/M

per million input tokens

OUTPUT

$0.378/M

per million output tokens

BLENDED 70/30

$0.290/M

default reference rate

CONTEXT

131K

131,072 tokens

What it's good at

Frontier quality, open weights
Aggressive pricing
Strong code & math
MoE architecture

Typical use cases

Self-hosted frontier inference
Cost-sensitive chat
Code generation

Benchmarks

vs. best public score

Scores inherited from DeepSeek-V3 — this is a hosted variant of the same open-weights model, so the underlying benchmark scores are identical.

MMLU88%

Multitask academic knowledge across 57 subjects.

GPQA Diamond59%

Graduate-level science questions, "Google-proof".

MATH90%

High-school competition math problems.

HumanEval89%

Python function synthesis from docstrings.

SWE-bench Verified42%

Real GitHub issues solved end-to-end.

LMArena Elo1318 Elo

Crowd-sourced head-to-head preference Elo rating.

Hand-curated from each provider's published reports and public leaderboards. Methodology varies across sources — treat as directional rather than authoritative.