DeepSeek-R1-0528

685B parameters (8 of 256 experts per token), text in, text out, 160K tokens of context. 5 providers serve it from $0.50 per million input tokens.

Verified 2026-10-25 Rank #41 in LLMs · score 109.2 method v1 Sources 4 Downloads 30d 165.2k · #266 of 619 tracked

Can I run it?411GB at INT4Estimateneeds a multi-GPU nodemethod

What does it cost?$0.50/1M incheapest of 4DeepInfrafp42026-10-25

Can I use it?MITPermissivelicence file2026-10-24

Does it work?71.4on Aider Polyglot#1 of 13 measuredleaderboard2025-06-06

Do people use it?165.2kdownloads, 30 days#266 of 619 we track · top 43%+4 likes, downloads −5% this week2,490 likesHugging Face2026-10-24

Worth reading

Picked by rule from 41 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Most discussed · 1,843 HN pts

DeepSeek-R1

github.com 2025-01-20 1,843 HN pts · 663 comments 2.1k words about the DeepSeek R1 family family

Latest · 6 months ago

ZAYA1-8B matches DeepSeek-R1 on math with less than 1B active parameters

Zyphra released ZAYA1-8B, a model that matches DeepSeek-R1 on math benchmarks while using less than 1 billion active parameters.

The model was trained entirely on a 1,024 node AMD cluster built with IBM using AMD Pensando Pollara interconnect.
Zyphra trained ZAYA1-8B end to end on AMD Instinct MI300X GPUs for pretraining, midtraining, and supervised fine-tuning.

firethering.com 2026-05-07 117 HN pts · 56 comments 1.3k words about the DeepSeek R1 family family

25 best of 41 tracked

The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

PaperDeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RLarxiv.org · 2025-01-25 · 1,351 HN pts
ArticlePromising results from DeepSeek R1 for codesimonwillison.net · 2025-01-28 · 979 HN pts
ArticleRun DeepSeek R1 Dynamic 1.58-bitunsloth.ai · 2025-01-28 · 767 HN pts
ArticleOpen-R1: an open reproduction of DeepSeek-R1huggingface.co · 2025-01-28 · 394 HN pts
CodeDeepseek R1 Distill 8B Q40 on 4 x Raspberry Pi 5github.com · 2025-02-15 · 306 HN pts
ArticleHow to Run DeepSeek R1 671B Locally on a $2000 EPYC Serverdigitalspaceport.com · 2025-02-01
CodeDeepSeek-R1-671B-Q4_K_M with 1 or 2 Arc A770 on Xeongithub.com · 2025-03-06
ArticleThe Illustrated DeepSeek-R1newsletter.languagemodels.co · 2025-01-27 · 578 HN pts
ArticleDeepSeek R1 Is Now Available on Azure AI Foundry and GitHubazure.microsoft.com · 2025-01-29 · 109 HN pts
DiscussionComplete hardware and software setup for running Deepseek-R1 locallytwitter.com · 2025-01-29 · 250 HN pts
CodeTinyZero: Reproduction of DeepSeek R1 Zero in countdown and multiplication tasksgithub.com · 2025-01-25
CodeOpen Reproduction of DeepSeek-R1github.com · 2025-01-27
ArticleOfficial DeepSeek R1 Now on Ollamaollama.com · 2025-01-20
ArticleBerkeley Researchers Replicate DeepSeek R1's Core Tech for Just $30: A Small Modxyzlabs.substack.com · 2025-01-28
CodeImplementing DeepSeek R1's GRPO algorithm from scratchgithub.com · 2025-04-13
ArticleMini-R1: Reproduce DeepSeek R1 "Aha Moment"philschmid.de · 2025-01-31
DiscussionDeepSeek R1 671B running on 2 M2 Ultras faster than reading speedtwitter.com · 2025-01-29
ArticleHow to run DeepSeek R1 locallyworkos.com · 2025-01-29
ArticleDeepSeek R1 0528 — Architecture, Evals & Inference Performance | InferenceXinferencex.semianalysis.com · undated
ArticleDeepSeek R1 0528 - Intelligence, Performance & Price ...artificialanalysis.ai · undated
ArticleHow to Run DeepSeek R1 Distilled Reasoning Models on RyzenAI and Radeon GPUsguru3d.com · 2025-02-02
ArticleA new, faster DeepSeek R1-0528 variant appears from German labventurebeat.com · 2025-07-05
ArticleDeepSeek-R1-0528 Did Not Have a Momentthezvi.substack.com · 2025-06-06

Show all 23

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

4 providers from $0.50

4 providers with a live price, from $0.50 to $0.70 per million input tokens; the median is $0.54. One dot per provider, cheapest in teal.

#	Cheapest 5	In /1M	Out /1M	Context	Quant	Seen
1	DeepInfra	$0.50	$2.15	160K	fp4	Oct 25
2	SiliconFlow	$0.50	$2.18	160K	fp8	Oct 24
3	StreamLake	$0.57	$2.29	125K	unknown	Oct 24
4	Novita	$0.70	$2.50	160K	—	Oct 25
5	Featherless	—	—	—	—	Oct 25

All 5 providers →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Widely served (5 API providers); 47 quantized builds for local use; a big base for builders (79 fine-tunes, adapters and merges).

As a service5DeepInfra, Featherless, Novita, SiliconFlow +1
On own hardware47quantized builds
Built on7959 fine-tunes · 16 adapters · 4 merges
In apps100+Hugging Face Spaces

Top Spaces running it: umint/ai ♥ 258 · hadadrjt/ai ♥ 86 · ginigen/deepseek-r1-0528-API ♥ 62 · linusorii/deepsite-v2 ♥ 61

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Needs an 8×80 GB node even at INT4

About 411 GB at INT4, 821 GB at FP8, 1.64 TB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.

Estimate params × bytes per param × 1.2 · how we estimate

How it scores

Third-party measurements only; author-reported numbers stay in Reference, marked as such.

#1 of 13 on Aider Polyglot so far

71.4. Each tick is a model on this site with a measured Aider Polyglot row; author-reported numbers stay in Reference. Percent (pass_rate_2). edit_format=diff

Source · leaderboard · 2025-06-06

The story so far

Weights, every dated article, and repo commits on one rail.

Most coverage came in the first two weeks

Teal is the weights (2025-05-28), amber triangles are 10 dated articles (each links to it), grey ticks are repo commits.

What people built on it

126 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

126 derivatives on the Hub, led by 59 fine-tunes

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

GGUF unsloth/DeepSeek-R1-0528-GGUF for llama.cpp, Ollama, LM Studio 474.7k downloads MLX mlx-community/DeepSeek-R1-0528-3bit for Apple silicon 6.3k downloads AWQ QuixiAI/DeepSeek-R1-0528-AWQ for vLLM, SGLang 28.5k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

24 rows

Benchmarks 24 rows

measured third-party leaderboard row (2)author-reported number from the model card or eval files (22)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.

Benchmark	Value	Setting / notes	Reported by	Date
measured Aider Polyglot · Write code with agents	71.4 percent (pass_rate_2)	edit_format=diff	leaderboard	2025-06-06
measured LiveCodeBench · Write code with agents	84.36 percent pass@1 (mean over questions)	all 1055 questions in file, 2023-05-08..2025-04-07	leaderboard	2025-04-07
author-reported AIME 2024 · Reason through hard problems	91.4 pass@1	DeepSeek R1 0528	model author	2026-10-07
author-reported AIME 2024 · Reason through hard problems	86	DeepSeek-R1-0528-Qwen3-8B	model author	2026-10-07
author-reported AIME 2025 · Reason through hard problems	87.5 pass@1	DeepSeek R1 0528	model author	2026-10-07
author-reported AIME 2025 · Reason through hard problems	76.3	DeepSeek-R1-0528-Qwen3-8B	model author	2026-10-07
author-reported Aider Polyglot · Write code with agents	71.6	DeepSeek R1 0528	model author	2026-10-07
author-reported BFCL · Call tools and APIs	37	DeepSeek R1 0528	model author	2026-10-07
author-reported CNMO 2024 (Pass@1)	86.9 pass@1	DeepSeek R1 0528	model author	2026-10-07
author-reported Codeforces · Write code with agents	1930 rating	DeepSeek R1 0528	model author	2026-10-07
author-reported FRAMES (Acc.)	83	DeepSeek R1 0528	model author	2026-10-07
author-reported GPQA Diamond · Reason through hard problems	81 pass@1	DeepSeek R1 0528	model author	2026-10-07
author-reported GPQA Diamond · Reason through hard problems	61.1	DeepSeek-R1-0528-Qwen3-8B	model author	2026-10-07
author-reported HMMT · Reason through hard problems	79.4 pass@1	DeepSeek R1 0528	model author	2026-10-07
author-reported HMMT · Reason through hard problems	61.5	DeepSeek-R1-0528-Qwen3-8B	model author	2026-10-07
author-reported Humanity's Last Exam · Reason through hard problems	17.7 pass@1	DeepSeek R1 0528	model author	2026-10-07
author-reported LiveCodeBench · Write code with agents	73.3 pass@1	DeepSeek R1 0528	model author	2026-10-07
author-reported LiveCodeBench · Write code with agents	60.5	DeepSeek-R1-0528-Qwen3-8B	model author	2026-10-07
author-reported MMLU · Reason through hard problems	93.4	DeepSeek R1 0528	model author	2026-10-07
author-reported MMLU-Pro · Reason through hard problems	85	DeepSeek R1 0528	model author	2026-10-07
author-reported SWE-bench Verified · Write code with agents	57.6	DeepSeek R1 0528	model author	2026-10-07
author-reported SimpleQA · Reason through hard problems	27.8	DeepSeek R1 0528	model author	2026-10-07
author-reported tau-bench · Call tools and APIs	53.5 pass@1	DeepSeek R1 0528 (Airline)	model author	2026-10-07
author-reported tau-bench · Call tools and APIs	63.9 pass@1	DeepSeek R1 0528 (Retail)	model author	2026-10-07

Use-case evidence

5 use cases

What it is good at 5 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

22 fields

Listed in LLMs · MoE

Facts

DeepSeek-R1-0528 has 685B parameters. (safetensors metadata, observed 2026-10-24)
DeepSeek-R1-0528 is a mixture-of-experts model with 256 experts, routing 8 per token. (config.json, observed 2026-10-24)
Context length is 163,840 tokens (160K). (config.json max_position_embeddings, observed 2026-10-24)
License: MIT (friction: none). Permissive. (license file, observed 2026-10-24)
Created 2025-05-28, last modified 2025-05-29. (HF API, observed 2026-10-24)

Pipeline tag	text-generation
Library	transformers
Architectures	DeepseekV3ForCausalLM
Model type	deepseek_v3
Dtype	bfloat16
Params by dtype	BF16: 3,918,786,560 · F32: 41,555,600 · F8_E4M3: 680,571,043,840
Size band	XXL
MoE	yes
Experts	256
Experts per token	8
Context length	163,840 (config.json max_position_embeddings)
Papers	arXiv:2501.12948
License	MIT (mit)
Hub snapshot	2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · DeepseekV3ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · DeepseekV3ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for DeepSeek-R1-0528: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)	1.64 TB · Cluster (≥640 GB)
FP8 (1 byte/param)	821 GB · Cluster (≥640 GB)
INT4 (0.5 bytes/param)	411 GB · Multi-GPU node (<640 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.

All variants

40 variants

Lineage

Original release (shipped quantized by the lab) Family: deepseek-ai/DeepSeek-R1 126 derivatives on the Hub

finetune: 59quantized: 47merge: 4adapter: 16

Top variants (40 shown, ranked by downloads)

Variant	Relation	Quant	Params	Band	License	Author	Likes	Downloads
nvidia/DeepSeek-R1-0528-NVFP4-v2	finetune	NVFP4	394B	XXL	mit	nvidia	27	5.4M
unsloth/DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	mit	unsloth	202	474.7k
nvidia/DeepSeek-R1-0528-NVFP4	finetune	NVFP4	397B	XXL	mit	nvidia	48	443.6k
amd/DeepSeek-R1-0528-MXFP4	quantized	MXFP4	356B	XXL	mit	amd	2	197.8k
amd/DeepSeek-R1-0528-MXFP4-v2	quantized	MXFP4	350B	XXL	mit	amd	0	149.8k
amd/DeepSeek-R1-0528-MXFP4-MTP-MoEFP4	quantized	MXFP4	350B	XXL	mit	amd	0	68.8k
bartowski/deepseek-ai_DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	mit	bartowski	11	58.2k
ubergarm/DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	mit	ubergarm	53	32.8k
tngtech/DeepSeek-TNG-R1T2-Chimera	merge	—	684B	XXL	mit	tngtech	287	21k
QuixiAI/DeepSeek-R1-0528-AWQ	quantized	AWQ	671B	XXL	mit	QuixiAI	19	28.5k
unsloth/DeepSeek-R1-0528-BF16	finetune	-BF16	684B	XXL	mit	unsloth	10	23.5k
RedHatAI/DeepSeek-R1-0528-quantized.w4a16	quantized	-quant	676B	XXL	mit	RedHatAI	13	14k
DevQuasar/deepseek-ai.DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	—	DevQuasar	7	10.5k
unsloth/DeepSeek-R1-0528	quantized	—	685B	XXL	mit	unsloth	15	9.5k
QuantTrio/DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Medium	quantized	GPTQ	912B	XXL	mit	QuantTrio	1	7.3k
mlx-community/DeepSeek-R1-0528-3bit	quantized	—	671B	XXL	mit	mlx-community	0	6.3k
lmstudio-community/DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	mit	lmstudio-community	9	5.2k
QuantTrio/DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Compact	quantized	GPTQ	847B	XXL	mit	QuantTrio	5	5.4k
amd/DeepSeek-R1-0528-MXFP4-ASQ	quantized	MXFP4	342B	XXL	mit	amd	1	5k
dyadya-sam/DeepSeek-R1-0528-GGUF-Q4_K_XL	quantized	GGUF	—	—	mit	dyadya-sam	0	3.4k
mlx-community/DeepSeek-R1-0528-5bit	quantized	—	671B	XXL	mit	mlx-community	3	1.8k
adamo1139/DeepSeek-R1-0528-AWQ	quantized	AWQ	671B	XXL	mit	adamo1139	4	1.6k
anikifoss/DeepSeek-R1-0528-DQ4_K_R4	quantized	GGUF	—	—	mit	anikifoss	9	753
QuantTrio/DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Lite	quantized	GPTQ	721B	XXL	mit	QuantTrio	2	963
moxin-org/DeepSeek-R1-0528-Moxin-GGUF	quantized	GGUF	—	—	mit	moxin-org	5	651
kaxap/mlx-DeepSeek-R1-0528-2bit	quantized	mlx	671B	XXL	mit	kaxap	0	860
AmpereComputing/deepseek-r1-0528-gguf	quantized	gguf	—	—	—	AmpereComputing	0	790
kaxap/mlx-DeepSeek-R1-0528-4bit	quantized	mlx	671B	XXL	mit	kaxap	0	748
bobchenyx/DeepSeek-R1-0528-HFK-GGUF	quantized	GGUF	—	—	mit	bobchenyx	0	721
NOYOUllm2/fortinet-lora-gguf	quantized	gguf	—	—	mit	NOYOUllm2	1	652
bullerwins/DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	mit	bullerwins	0	606
tachyphylaxis/DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	mit	tachyphylaxis	0	587
bobchenyx/DeepSeek-R1-0528-MLA-GGUF	quantized	GGUF	—	—	mit	bobchenyx	1	517
ArtusDev/deepseek-ai_DeepSeek-R1-0528-GGUF	quantized	GGUF	—	—	mit	ArtusDev	0	520
TVS759/DeepSeek-TNG-R1T2-Chimera	merge	—	684B	XXL	mit	TVS759	0	439
mmbela/DeepSeek-R1-0528-optimized-for-512Gb-GGUF	quantized	GGUF	—	—	mit	mmbela	1	352
Jinx-org/Jinx-DeepSeek-R1-0528	finetune	—	684B	XXL	apache-2.0	Jinx-org	6	73
amd/DeepSeek-R1-0528-ptpc	quantized	—	671B	XXL	mit	amd	0	367
ling1000T/DeepSeek-R1-0528-gguf	quantized	gguf	—	—	mit	ling1000T	1	298
gghfez/DeepSeek-R1-0528-256x21B-BF16	finetune	-BF16	—	—	mit	gghfez	0	320

Repository changelog

13 commits

Repository changelog 13 commits

Date	Commit	SHA
2025-05-29	Release DeepSeek-R1-0528	4236a6a
2025-05-29	Update .gitattributes	54b97b8
2025-05-28	Add files using upload-large-folder tool	174da7f
2025-05-28	Add files using upload-large-folder tool	f90ddc6
2025-05-28	Add files using upload-large-folder tool	8f1a7bc
2025-05-28	Add files using upload-large-folder tool	8504b97
2025-05-28	Add files using upload-large-folder tool	d9d887f
2025-05-28	Add files using upload-large-folder tool	894d8c6
2025-05-28	Add files using upload-large-folder tool	ecba5ba
2025-05-28	Add files using upload-large-folder tool	4823201
2025-05-28	Add files using upload-large-folder tool	2fb3b9e
2025-05-28	Add files using upload-large-folder tool	1162836
2025-05-28	initial commit	f3602d5

All providers

5 providers · 5 endpoints

Where to run 5 providers · 5 endpoints

Hosted endpoints observed for DeepSeek-R1-0528, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.50/M.

Provider	Model ref	Context	Quant	$/M in	$/M out	tok/s	TTFT	Tools	Status	Observed
DeepInfra hf, deepinfra, openrouter	deepseek-ai/DeepSeek-R1-0528	160K	fp4	$0.50	$2.15	23.1	726 ms	yes	live	2026-10-25
SiliconFlow openrouter	deepseek/deepseek-r1-0528	160K	fp8	$0.50	$2.18	—	—	yes	live	2026-10-24
StreamLake openrouter	deepseek/deepseek-r1-0528	125K	unknown	$0.57	$2.29	—	—	—	live	2026-10-24
Novita hf	deepseek/deepseek-r1-0528	160K	—	$0.70	$2.50	25.8	914 ms	yes	live	2026-10-25
Featherless hf	deepseek-ai/DeepSeek-R1-0528	—	—	—	—	—	—	—	live	2026-10-25

Articles and coverage

41 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Analysis9

ZAYA1-8B matches DeepSeek-R1 on math with less than 1B active parameters

firethering.com2026-05-07HN front pageabout the DeepSeek R1 familyHN 117 pts · 56 comments1,293 words

Zyphra released ZAYA1-8B, a model that matches DeepSeek-R1 on math benchmarks while using less than 1 billion active parameters. The 8.4B parameter mixture of experts model activates only 760M parameters at inference and employs Markovian RSA to scale performance with compute budget. It requires a specific vLLM fork for local deployment and is available under an Apache 2.0 license.written, reviewed, cited

The model was trained entirely on a 1,024 node AMD cluster built with IBM using AMD Pensando Pollara interconnect.
Zyphra trained ZAYA1-8B end to end on AMD Instinct MI300X GPUs for pretraining, midtraining, and supervised fine-tuning.
The model requires Zyphra's own fork of vLLM to run properly because the standard vLLM install won't work.

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL

arxiv.org2025-01-25authorityHN front pageabout the DeepSeek R1 familyHN 1,351 pts · 1,056 comments371 words

Promising results from DeepSeek R1 for code

simonwillison.net2025-01-28authorityHN front pageabout the DeepSeek R1 familyHN 979 pts · 746 comments345 words

A PR for llama.cpp delivers x2 speed for WASM by optimizing SIMD instructions for qX_K_q8_K and qX_0_q8_0 dot product functions. Surprisingly, 99% of the code in this PR is written by DeekSeek-R1. The model spent 3-5 minutes thinking about each prompt to generate the optimization.written, reviewed, cited

DeepSeek-R1

github.com2025-01-20authorityHN front pageabout the DeepSeek R1 familyHN 1,843 pts · 663 comments2,107 words

Run DeepSeek R1 Dynamic 1.58-bit

unsloth.ai2025-01-28authorityHN front pageabout the DeepSeek R1 familyHN 767 pts · 332 comments

Open-R1: an open reproduction of DeepSeek-R1

huggingface.co2025-01-28authorityHN front pageabout the DeepSeek R1 familyHN 394 pts · 234 comments890 words

Deepseek R1 Distill 8B Q40 on 4 x Raspberry Pi 5

github.com2025-02-15authorityHN front pageabout the DeepSeek R1 familyHN 306 pts · 156 comments766 words

The Illustrated DeepSeek-R1

newsletter.languagemodels.co2025-01-27HN front pageabout the DeepSeek R1 familyHN 578 pts · 118 comments1,715 words

DeepSeek R1 Is Now Available on Azure AI Foundry and GitHub

azure.microsoft.com2025-01-29authorityHN front pageabout the DeepSeek R1 familyHN 109 pts · 49 comments

Community1

Complete hardware and software setup for running Deepseek-R1 locally

twitter.com2025-01-29HN front pageabout the DeepSeek R1 familyHN 250 pts · 192 comments

All coverage tracked for this model (41) — append-only, nothing is ever removed

Title	Source	Published	First seen	Signals
DeepSeek R1 0528 — Architecture, Evals & Inference Performance \| InferenceX	inferencex.semianalysis.com	—	2026-10-07	exa
DeepSeek R1 0528 - Intelligence, Performance & Price ...	artificialanalysis.ai	—	2026-10-07	exa
DeepSeek \| DeepSeek R1 0528 \| Kaggle	kaggle.com	—	2026-10-07	exa
DeepSeek R1 0528 Deep Dive	runpod.io	—	2026-10-07	exa
DeepSeek: R1 0528 - AI Model Details & Benchmarks	benchable.ai	—	2026-10-07	exa
ZAYA1-8B matches DeepSeek-R1 on math with less than 1B active parameters	firethering.com	2026-05-07	2026-10-07	hn
A new, faster DeepSeek R1-0528 variant appears from German lab	venturebeat.com	2025-07-05	2026-10-07	hn
DeepSeek R1 0528 scored 71% on the aider polyglot coding benchmark (3rd)	twitter.com	2025-06-09	2026-10-07	hn
DeepSeek-R1-0528 Did Not Have a Moment	thezvi.substack.com	2025-06-06	2026-10-07	hn
The 'Minor Upgrade' That's Anything But: DeepSeek R1-0528 Deep Dive	runpod.ghost.io	2025-05-31	2026-10-07	exa
DeepSeek-R1-0528: A Detailed Review of its AI Coding Performance & Latency \| ForgeCode	forgecode.dev	2025-05-30	2026-10-07	exa
DeepSeek-R1-0528 Released	api-docs.deepseek.com official	2025-05-29	2026-10-07	hn
DeepSeek-R1-0528 is here	twitter.com	2025-05-29	2026-10-07	hn
Deepseek-R1-0528 Dynamic 1-bit	unsloth.ai	2025-05-29	2026-10-07	hn
DeepSeek-R1-0528 is now live on Hyperbolic	app.hyperbolic.xyz	2025-05-28	2026-10-07	hn
DeepSeek R1-0528: The Update That Quietly Closed the Gap With o3 — Redeep	redeep.blog	2025-05-28	2026-10-07	exa
DeepSeek R1 (0528) Review — Benchmarks, Pricing & AI Panel Verdict \| TopReviewed.ai	topreviewed.ai	2025-05-27	2026-10-07	exa
Implementing DeepSeek R1's GRPO algorithm from scratch	github.com	2025-04-13	2026-10-07	hn
DeepSeek-R1-671B-Q4_K_M with 1 or 2 Arc A770 on Xeon	github.com	2025-03-06	2026-10-07	hn
Deepseek R1 Distill 8B Q40 on 4 x Raspberry Pi 5	github.com	2025-02-15	2026-10-07	hn
Experience the DeepSeek R1 Distilled 'Reasoning' Models on Ryzen AI and Radeon	community.amd.com	2025-02-04	2026-10-07	hn
How to Run DeepSeek R1 Distilled Reasoning Models on RyzenAI and Radeon GPUs	guru3d.com	2025-02-02	2026-10-07	hn
How to Run DeepSeek R1 671B Locally on a $2000 EPYC Server	digitalspaceport.com	2025-02-01	2026-10-07	hn
Running DeepSeek R1 Models Locally on NPU	blogs.windows.com	2025-02-01	2026-10-07	hn
How to run 1.58bit DeepSeek R1 with Open WebUI	docs.openwebui.com	2025-01-31	2026-10-07	hn
Mini-R1: Reproduce DeepSeek R1 "Aha Moment"	philschmid.de	2025-01-31	2026-10-07	hn
DeepSeek R1 is far better than OpenAI o1 at finding bugs in pull requests	greptile.com	2025-01-30	2026-10-07	hn
DeepSeek R1 Is Now Available on Azure AI Foundry and GitHub	azure.microsoft.com	2025-01-29	2026-10-07	hn
How to run DeepSeek R1 locally	workos.com	2025-01-29	2026-10-07	hn
Complete hardware and software setup for running Deepseek-R1 locally	twitter.com	2025-01-29	2026-10-07	hn
DeepSeek R1 671B running on 2 M2 Ultras faster than reading speed	twitter.com	2025-01-29	2026-10-07	hn
Berkeley Researchers Replicate DeepSeek R1's Core Tech for Just $30: A Small Mod	xyzlabs.substack.com	2025-01-28	2026-10-07	hn
Promising results from DeepSeek R1 for code	simonwillison.net	2025-01-28	2026-10-07	hn
Run DeepSeek R1 Dynamic 1.58-bit	unsloth.ai	2025-01-28	2026-10-07	hn
Open-R1: an open reproduction of DeepSeek-R1	huggingface.co	2025-01-28	2026-10-07	hn
The Illustrated DeepSeek-R1	newsletter.languagemodels.co	2025-01-27	2026-10-07	hn
Open Reproduction of DeepSeek-R1	github.com	2025-01-27	2026-10-07	hn
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL	arxiv.org	2025-01-25	2026-10-07	hn
TinyZero: Reproduction of DeepSeek R1 Zero in countdown and multiplication tasks	github.com	2025-01-25	2026-10-07	hn
Official DeepSeek R1 Now on Ollama	ollama.com	2025-01-20	2026-10-07	hn
DeepSeek-R1	github.com	2025-01-20	2026-10-07	hn

Papers

1 paper

Papers 1

arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning this model peer-reviewed · reported

DeepSeek-AI, Daya Guo, Dejian Yang +197 · 2025-01-22 · Nature volume 645, pages 633-638 (2025) · PDF · HF paper page (462 upvotes) · evidence

Similar models

3 models

Similar models

GLM-5.3 · 753B
same size band (XXL), also MoE, same task (text-generation), more providers
GLM-5.2 · 753B
same size band (XXL), also MoE, same task (text-generation), more providers, permissive license
DeepSeek-V3 · 684B
same size band (XXL), also MoE, same task (text-generation), fewer providers

Sources

4 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:32 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 14:59 UTC · Last verified 2026-10-25

Follow DeepSeek-R1-0528

Watch this model

/feeds/models/deepseek-ai--DeepSeek-R1-0528.xml(RSS 2.0) · every change to DeepSeek-R1-0528
/feeds/families/deepseek-ai--DeepSeek-R1.xml(RSS 2.0) · every change across the DeepSeek-R1 family

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.