DeepSeek

30 models tracked · known for DeepSeek-R1, DeepSeek-R1, DeepSeek-V4-Flash-0731

Models tracked

162

Benchmark scores

2029

Best score

21%

Reproduced

34 of 162

What the record says

Claims drawn from cited facts, not live model generation.

3 claims

Lab Summary

DeepSeek is an artificial intelligence laboratory.

1 cited fact

Portfolio

DeepSeek has shipped 30 models. This tracked footprint is centered solely on model development, signaling a lab whose portfolio is built around shipped models.

3 cited facts

Evidence mix

162 benchmark scores across this lab’s models, tagged by evidence status.

reproduced34self-reported69unverified59

Models

Status

Sort

Model	Best score	Claim status
DeepSeek-R1	2029	contradicted
DeepSeek-R1-Distill-Qwen-32B	1691	self-reported
DeepSeek-R1-Distill-Llama-70B	1633	self-reported
DeepSeek-R1-Distill-Qwen-14B	1481	self-reported
DeepSeek-R1-Distill-Llama-8B	1205	self-reported
DeepSeek-R1-Distill-Qwen-7B	1189	self-reported
DeepSeek-V3	1134	contradicted
DeepSeek-R1-Distill-Qwen-1.5B	954	self-reported
DeepSeek-V4-Pro	96.66	independently reproduced
DeepSeek-R1 (May 2025)	96.64	independently reproduced
DeepSeek V4 Flash 0731	94.44	independently reproduced
DeepSeek-V2 (MoE-236B, May 2024)	89.6	self-reported
DeepSeek-V3.2	87.81	independently reproduced
DeepSeek-V3.2-Exp	83.3	unverified
DeepSeek-V3 (Mar 2025)	77	independently reproduced
DeepSeek-Coder-V2-Lite-Base	67.1	self-reported
DeepSeek Coder 33B	35.4	self-reported
DeepSeek Coder 6.7B	21.3	self-reported
DeepSeek Coder 1.3B	6.6	self-reported
DeepSeek-V4-Flash-0731	—	unverified
DeepSeek-V4-Flash	—	unverified
DeepSeek-R1-0528-Qwen3-8B	—	unverified
DeepSeek-V3-0324	—	unverified
DeepSeek-V2-Lite-Chat	—	unverified
deepseek-coder-7b-instruct-v1.5	—	unverified
DeepSeek-R1-Distill-Qwen-14B	—	unverified

Benchmarks led

Where a model from DeepSeek holds the top score, with the benchmark’s integrity grade.

Source receipt

Source facts, citations, and refresh stamp for this record.

updated 2026-08-21

Source records: 139
Fact rows: 275
Fact types: 2
Sources: 3

Sources: curated_models, epoch_benchmarks, huggingface_models

DeepSeek

What the record says

Lab Summary

Portfolio

Evidence mix

Models

Benchmarks led

Source receipt

Follow the record

Other labs

Explore the record

DeepSeek

What the record says

Lab Summary

Portfolio

Evidence mix

Models

Benchmarks led

Source receipt

Follow the record

Other labs

Explore the record