Skip to content
← All models
DeepSeek

DeepSeek V3.2 (Thinking)

DeepSeek Open weights
80.7
Vibe Coding Index · #12 of 21
✓ Verified data — every figure is cited. See sources.
Quality #12/21
80.7
VCI
Price
—
$/M blended
Speed
—
tok/s
Latency
—
to first token

Suite radar

Radar hidden — some suite scores are unverified.

Suite scores

SWE-bench Verified
Real GitHub issues fixed end-to-end
—
Terminal-Bench
Agentic tasks in a real shell
—
LiveCodeBench
Fresh competitive-programming problems
81
↗

Where DeepSeek V3.2 (Thinking) sits

Its dot is ringed on the cost×quality and speed×quality frontiers.

↖ Best value Premium ↗ ↙ Budget Poor value ↘ 45 50 55 60 65 70 75 80 85 90 95 $0.32 $1 $3.16 $10 $31.62 BLENDED COST — $ / 1M TOKENS (LOG) VIBE CODING INDEX Claude Fable 5 Grok 4.5 MiniMax M3
Efficient frontier Alibaba Anthropic DeepSeek Google MiniMax Moonshot AI OpenAI xAI Hover, tap or focus a point · dashed lines = medians 10 model(s) hidden — data not yet verified
↖ Smart but slow Fast & smart ↗ ↙ Laggards Fast but shallow ↘ 45 50 55 60 65 70 75 80 85 90 95 0 48 96 144 192 240 OUTPUT SPEED — TOKENS / SECOND VIBE CODING INDEX Claude Fable 5 Grok 4.5 MiniMax M3 Qwen3.7 Max
Efficient frontier Alibaba Anthropic DeepSeek Google MiniMax Moonshot AI OpenAI xAI Hover, tap or focus a point · dashed lines = medians 10 model(s) hidden — data not yet verified
Compare DeepSeek V3.2 (Thinking) → Back to leaderboard