What is the latest version of Cerebras WSE-3?

The current shipped version of Cerebras WSE-3 is WSE-3, as of September 15, 2026.

How much does Cerebras WSE-3 cost?

Cerebras WSE-3 has no published flat entry price. Free tier / developer tier pricing available (contact/partner integrations)

What AI model does Cerebras WSE-3 use?

Cerebras WSE-3 runs primarily on Codex-Spark.

← Chips & Compute

Cerebras WSE-3

Name: Cerebras WSE-3
Author: Cerebras

Cerebras · #3 most active of 18 in Chips & Compute

Compare →

Single-wafer accelerators delivering the fastest published open-model token rates.

Trust score · Mixed

3 data points checked · 2026-09-12

Model stack: VerifiedEntry price: DisputedVersion: DisputedHow this works →

Current version

WSE-3

Entry cost

Free tier / developer tier pricing available (contact/partner integrations)

Changes / 30d

The short answers · verified September 15, 2026

What is the latest version of Cerebras WSE-3?: The current shipped version of Cerebras WSE-3 is WSE-3, as of September 15, 2026.
How much does Cerebras WSE-3 cost?: Cerebras WSE-3 has no published flat entry price. Free tier / developer tier pricing available (contact/partner integrations)
What AI model does Cerebras WSE-3 use?: Cerebras WSE-3 runs primarily on Codex-Spark.

Source: Cerebras — www.cerebras.ai · Reusable under CC BY 4.0 — cite Tomorrow

Capabilities

wafer-scale engine
record token/s on open models
CS-3 systems
training clusters
Wafer-scale AI inference
Cerebras PyTorch
ModelZoo
Low-level SDK
OpenAI API compatibility
Disaggregated Inference
Wafer-Scale Engine (58x larger than GPUs)
15x faster than GPUs
Cerebras PyTorch & ModelZoo support
Ultra-fast token generation
Low-latency inference
Model fine-tuning and training services
Wafer-Scale Engine
Ultra-fast AI Inference
Model fine-tuning and training
Custom model weights support
PyTorch support
58x larger than GPUs
World's largest AI-optimized processor
Model training and fine-tuning
Ultra-fast AI inference
Wafer-Scale Engine AI hardware
AI model training and fine-tuning
Drop-in OpenAI API compatibility
Inference API
PyTorch Training
Model fine-tuning
AI training
AI inference
Low-latency programming
AI Inference
Model Fine-tuning
Model Training
OpenAI API Compatibility
Training
Fine-tuning
Training and fine-tuning services
Dedicated queue priority
Custom model weights
AI Hardware / Chip
AI Training & Fine-tuning
Wafer-scale hardware acceleration
LLM Inference API
Model training
SDK
Developer support
AI Training
PyTorch Support
AI Hardware Accelerator
Inference Cloud API
World-record speeds
AWS Marketplace Integration
OpenRouter Integration
HuggingFace Integration
Vercel Integration
Training services
Fastest AI inference
Developer API
AWS Marketplace integration
OpenRouter integration
HuggingFace integration
Vercel integration
AI Accelerator
AWS Marketplace
AI Inference API
Model training services
Low latency
Inference
HPC applications
Up to 30x faster than GPUs
Dedicated on-premise deployment
Pytorch Support
inference
training
fine-tuning
api-access
on-premise
Developer SDK
Low-latency queue priority
GPU-free Wafer-Scale Engine
OpenAI-compatible API
Fine-tuning and training services
Developer Tier Pricing
Enterprise SLA
Low latency inference
Dedicated support
Up to 30x faster inference than GPU
Pre-training
PyTorch SDK
Low-latency token generation
OpenAI API compatible
Custom kernels
On-Premise Deployment
Cloud API

Pricing

Developer Tier
Free

Entry price over time

11 of 49 points reconstructed from Internet Archive captures (hollow dots) — dated by capture, not by when we first saw the page. View a capture

9/2/2025 · $15009/15/2026 · $0

Geek mode

Models under the hood

Codex-Spark
Cerebras
disclosed
Gemma
Google
inferred
Gemma-4-31B
Google
inferred
GLM-4.7
Zhipu
disclosed
GPT 5.6 Sol
OpenAI
disclosed
GPT-5.6-Sol-Ultrafast
OpenAI
disclosed
Kimi K2.6
Moonshot
disclosed
Llama
Meta
inferred
Llama / Qwen / GPT-OSS
open weights
disclosed
Qwen
Alibaba
inferred
WSE-3
Cerebras
disclosed

Context window

—

Public API

yes

Multi-model routing

yes

Stack signals

PyTorch

Shared model stack

Other tracked products running on the same foundation models — a quick read on how much of the catalog moves when one of these models changes.

Llama+1 aliasMeta · 10 products
GPT-5.6 Sol+1 aliasOpenAI · 7 products
QwenAlibaba · 7 products
GLM-4.7Zhipu AI · 4 products
Kimi K2.6Moonshot · 3 products
GemmaGoogle · 2 products
- Ollama
- Cerebras WSE-3
Codex-SparkCerebras · 1 product
- Cerebras WSE-3
Gemma-4-31BGoogle · 1 product
- Cerebras WSE-3

green disclosed · amber inferred · grey unattributed. Aliases are folded into one model; provider concentration counts only models with an established vendor.

Reported scale

Reference data. Each figure is whatever the source actually said — weekly users, downloads, revenue run-rate — with its own definition and date. These are not comparable between tools and are never used to rank anything.

No public usage figure on record for this product.

Also in Chips & Compute

All Chips & Compute

Tenstorrent Blackhole

Tenstorrent

Open RISC-V AI hardware

Version: Blackhole
Cost: from $999/unit
Model: Tensix cores

Google TPU Ironwood

Google

Inference-first TPU pods

Version: Ironwood (7th generation)
Cost: from $5.4/unit
Model: Gemini

NVIDIA Blackwell

NVIDIA

GB200/B200 rack-scale AI systems

Version: Blackwell
Cost: Pricing is not publicly listed (enterprise hardware architecture/system)
Model: Blackwell

AWS Trainium2

AWS

Cloud-native training silicon

Version: Trainium2
Cost: Pricing not published; available via AWS EC2 instance pricing.
Model: Neuron compiler

Intel Gaudi 3

Intel

Cost-focused training and inference accelerator.

Version: 1.24.1
Cost: Pricing not published. Available via Intel Tiber AI Cloud or OEM platforms.
Model: Accelerator

SambaNova SN40L

SambaNova

Dataflow chips for model serving

Version: SN40L
Cost: Free tier available. Paid plans include Developer (Pay-as-you-go) and Enterprise (Subscription-based).
Model: DeepSeek-V3.1

Elsewhere on the same models

Products in other segments that name one of Cerebras WSE-3's models in their stack.

Fireworks AI

Low-latency serving for open models and custom fine-tunes.

Version: 2025
Cost: Per-token usage
Model: Llama

Kimi

Moonshot AIChina

Long-context assistant and agentic K-series models.

Version: Kimi K3
Cost: API access with pay-as-you-go pricing based on token consumption. File APIs are currently free.
Model: Kimi K2.6

Ollama

Local model runner that made desktop inference trivial.

Version: 0.x
Cost: Free
Model: Gemma

Qwen-Image

AlibabaChina

Open-weight image generation and editing from the Qwen line.

Version: Qwen-Image
Cost: Pricing not specified in the provided documentation.
Model: Qwen

Together AI

Open-model inference and fine-tuning at frontier speed.

Version: 2025
Cost: Per-token, from $0.06/M
Model: Llama

Change history

Sep 15, 2026first seencapability
New capabilities: On-Premise Deployment, Cloud API
96 tracked98 tracked · +On-Premise Deployment, Cloud API
source
Sep 14, 2026first seenpricing
Entry price decreased to $0/mo
$10$0
source
Sep 14, 2026first seencapability
New capabilities: OpenAI API compatible, Custom kernels
94 tracked96 tracked · +OpenAI API compatible, Custom kernels
source
Sep 12, 2026first seencapability
New capabilities: PyTorch SDK, Low-latency token generation
92 tracked94 tracked · +PyTorch SDK, Low-latency token generation
source
Sep 11, 2026first seencapability
New capabilities: Up to 30x faster inference than GPU, Pre-training
90 tracked92 tracked · +Up to 30x faster inference than GPU, Pre-training
source
Sep 10, 2026first seencapability
New capabilities: Low latency inference, Dedicated support
88 tracked90 tracked · +Low latency inference, Dedicated support
source
Sep 9, 2026first seencapability
New capabilities: Developer Tier Pricing, Enterprise SLA
86 tracked88 tracked · +Developer Tier Pricing, Enterprise SLA
source
Sep 8, 2026first seencapability
New capabilities: GPU-free Wafer-Scale Engine, OpenAI-compatible API, Fine-tuning and training services
83 tracked86 tracked · +GPU-free Wafer-Scale Engine, OpenAI-compatible API, Fine-tuning and training services
source
Sep 7, 2026first seencapability
New capabilities: Developer SDK, Low-latency queue priority
81 tracked83 tracked · +Developer SDK, Low-latency queue priority
source
Sep 6, 2026first seencapability
New capabilities: inference, training, fine-tuning
76 tracked81 tracked · +inference, training, fine-tuning, api-access, on-premise
source
Sep 5, 2026first seencapability
New capabilities: Pytorch Support
75 tracked76 tracked · +Pytorch Support
source
Sep 4, 2026first seencapability
New capabilities: Up to 30x faster than GPUs, Dedicated on-premise deployment
73 tracked75 tracked · +Up to 30x faster than GPUs, Dedicated on-premise deployment
source
Sep 3, 2026first seencapability
New capabilities: HPC applications
72 tracked73 tracked · +HPC applications
source
Sep 1, 2026first seencapability
New capabilities: Inference
71 tracked72 tracked · +Inference
source
Aug 30, 2026first seencapability
New capabilities: Low latency
70 tracked71 tracked · +Low latency
source
Aug 29, 2026first seenversion
Cerebras WSE-3 moved to WSE-3
CS-3WSE-3
source
Aug 28, 2026first seenversion
Cerebras WSE-3 moved to CS-3
WSE-3CS-3
source
Aug 28, 2026first seencapability
New capabilities: AI Inference API, Model training services
68 tracked70 tracked · +AI Inference API, Model training services
source
Aug 27, 2026first seenversion
Cerebras WSE-3 moved to WSE-3
CS-4WSE-3
source
Aug 26, 2026first seenversion
Cerebras WSE-3 moved to CS-4
WSE-3CS-4
source
Aug 26, 2026first seencapability
New capabilities: AI Accelerator, AWS Marketplace
66 tracked68 tracked · +AI Accelerator, AWS Marketplace
source
Aug 25, 2026first seencapability
New capabilities: Fastest AI inference, Developer API, AWS Marketplace integration
60 tracked66 tracked · +Fastest AI inference, Developer API, AWS Marketplace integration, OpenRouter integration, HuggingFace integration, Vercel integration
source
Aug 24, 2026first seencapability
New capabilities: Training services
59 tracked60 tracked · +Training services
source
Aug 23, 2026first seenversion
Cerebras WSE-3 moved to WSE-3
CS-4WSE-3
source
Aug 22, 2026first seenversion
Cerebras WSE-3 moved to CS-4
CS-3CS-4
source
Aug 22, 2026first seenmodel
Cerebras WSE-3 added Codex-Spark to its model stack
GLM-4.7, GPT 5.6 Sol, GPT-5.6-Sol-Ultrafast, Gemma, Gemma-4-31B, Kimi K2.6, Llama, Llama / Qwen / GPT-OSS, Qwen, WSE-3Codex-Spark, GLM-4.7, GPT 5.6 Sol, GPT-5.6-Sol-Ultrafast, Gemma, Gemma-4-31B, Kimi K2.6, Llama, Llama / Qwen / GPT-OSS, Qwen, WSE-3
source
Aug 22, 2026first seencapability
New capabilities: AWS Marketplace Integration, OpenRouter Integration, HuggingFace Integration
55 tracked59 tracked · +AWS Marketplace Integration, OpenRouter Integration, HuggingFace Integration, Vercel Integration
source
Aug 21, 2026first seenversion
Cerebras WSE-3 moved to CS-3
WSE-3CS-3
source
Aug 21, 2026first seencapability
New capabilities: AI Hardware Accelerator, Inference Cloud API, World-record speeds
52 tracked55 tracked · +AI Hardware Accelerator, Inference Cloud API, World-record speeds
source
Aug 20, 2026first seencapability
New capabilities: AI Training, PyTorch Support
50 tracked52 tracked · +AI Training, PyTorch Support
source
Aug 15, 2026first seenmodel
Cerebras WSE-3 added GPT 5.6 Sol, GPT-5.6-Sol-Ultrafast to its model stack
GLM-4.7, Gemma, Gemma-4-31B, Kimi K2.6, Llama, Llama / Qwen / GPT-OSS, Qwen, WSE-3GLM-4.7, GPT 5.6 Sol, GPT-5.6-Sol-Ultrafast, Gemma, Gemma-4-31B, Kimi K2.6, Llama, Llama / Qwen / GPT-OSS, Qwen, WSE-3
source
Aug 15, 2026first seencapability
New capabilities: Model training, SDK, Developer support
47 tracked50 tracked · +Model training, SDK, Developer support
source
Aug 14, 2026first seencapability
New capabilities: Wafer-scale hardware acceleration, LLM Inference API
45 tracked47 tracked · +Wafer-scale hardware acceleration, LLM Inference API
source
Aug 13, 2026first seencapability
New capabilities: AI Hardware / Chip, AI Training & Fine-tuning
43 tracked45 tracked · +AI Hardware / Chip, AI Training & Fine-tuning
source
Aug 12, 2026first seencapability
New capabilities: Training and fine-tuning services, Dedicated queue priority, Custom model weights
40 tracked43 tracked · +Training and fine-tuning services, Dedicated queue priority, Custom model weights
source
Aug 11, 2026first seencapability
New capabilities: Training, Fine-tuning
38 tracked40 tracked · +Training, Fine-tuning
source
Aug 10, 2026first seencapability
New capabilities: AI Inference, Model Fine-tuning, Model Training
34 tracked38 tracked · +AI Inference, Model Fine-tuning, Model Training, OpenAI API Compatibility
source
Aug 9, 2026first seenversion
Cerebras WSE-3 moved to WSE-3
CS-3WSE-3
source
Aug 9, 2026first seencapability
New capabilities: AI training, AI inference, Low-latency programming
31 tracked34 tracked · +AI training, AI inference, Low-latency programming
source
Aug 8, 2026first seenversion
Cerebras WSE-3 moved to CS-3
WSE-3CS-3
source

Subscribe to Cerebras WSE-3 changes

Cerebras WSE-3 compared

Straight head-to-head pages against the busiest products in Chips & Compute.

Cerebras WSE-3 vs Tenstorrent Blackhole Cerebras WSE-3 vs Google TPU Ironwood Cerebras WSE-3 vs NVIDIA Blackwell Cerebras WSE-3 vs AWS Trainium2 Cerebras WSE-3 vs Intel Gaudi 3 Cerebras WSE-3 vs SambaNova SN40L

How to cite this page

Free to cite and reuse under CC BY 4.0. Permalink: https://tomorrow.aliensquad.ai/tools/cerebras

APA

Tomorrow. (2026). Cerebras WSE-3 — version, pricing and model stack [Data set entry]. AlienSquad. Retrieved 2026-09-17, from https://tomorrow.aliensquad.ai/tools/cerebras

BibTeX

@misc{tomorrow-tools-cerebras,
  author       = {{Tomorrow}},
  title        = {Cerebras WSE-3 — version, pricing and model stack},
  year         = {2026},
  publisher    = {AlienSquad},
  howpublished = {\url{https://tomorrow.aliensquad.ai/tools/cerebras}},
  note         = {Accessed: 2026-09-17}
}