DeepSeekopen-weight
Released: 2024-12-26

DeepSeek-V3

Ultra-cost-efficient open-weight MoE architecture offering near-frontier intelligence at a fraction of API costs.

Intelligence Index
82.4/ 100
Calibrated multi-domain composite
Throughput Speed
62tok/sec
Output streaming throughput
Time To First Token
410ms
Initial chunk server latency
LiveBench Contam-Free
67.8%
September 2026 suite score

LiveBench Multi-Domain Evaluation Breakdown

Monthly refreshed contamination-free test sets
Logical Reasoning81.9%
Coding & Repo Repair69.2%
Mathematics (AIME)78.9%
Data Analysis & Tables74.2%
Language & Comprehension80.5%
Instruction Following87.5%

Sourced Benchmark Evaluations (5)

Verified performance across authoritative benchmarks with provenance tracking.
BenchmarkStatusScoreTrustDateSource TypeProvenance
StructFlowBenchNearing Saturation96%
75
2025-02-20independentSource ↗
ARC-AGI-2Nearing Saturation63.9%
75
2025-01-23independentSource ↗
SWE-bench VerifiedSaturated52.6%
25
2025-01-04independentSource ↗
IFEvalDeprecated87.5%
25
2024-12-30independentSource ↗
BIG-Bench HardSaturated87.5%
25
2024-12-26vendor-reportedSource ↗
Local Hardware Execution (Ollama)Hardware Compatible
# 1. Pull and execute model locally with Ollama
ollama run deepseek-v3

# 2. Or invoke via local OpenAI-compatible endpoint
curl http://localhost:11434/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v3",
    "messages": [{"role": "user", "content": "Analyze reasoning chains on AIME 2026"}]
  }'

Evaluation & Sourcing Notes

Scores listed for DeepSeek-V3 represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.

All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.