Cerebras

Wafer-scale inference, the fastest tokens per second available

cerebras.ai inference-docs.cerebras.ai

Overall in llm apis

17of 18 · score 56.6

Data coverage

93%

25 of 27 fields sourced

Best for

•Reasoning models where thinking-token latency is the bottleneck
•Interactive tools that feel broken at 50 tokens per second
•Benchmarking how much of your UX problem is actually latency

Strong on

+Zero retention Yes 1/18

Weak on

−Context window 131,072 tokens 12/14
Better: Anthropic: 1,000,000 tokens
−Image input No 17/18
Better: Anthropic: Yes
−Max output 40,000 tokens 9/10
Better: xAI: Unlimited
−API since 2024 16/18
Better: OpenAI: 2020

vs Fireworks AI

•Behind on Image input: No vs Partial
•Behind on Audio in/out: No vs Partial

How Cerebras scores among 18 llm apis25 of 27 fields known

Capability7

Flagship model

Curated open-weight models (Qwen3, GLM, Llama, GPT-OSS)

inference-docs.cerebras.ai · 2026-01-15

Context window

131,072 tokens

12/14

Leader: Anthropic · 1,000,000 tokens

inference-docs.cerebras.ai · 2026-11-07

Max output

40,000 tokens

9/10

Leader: xAI · Unlimited

inference-docs.cerebras.ai · 2026-11-07

Image input

17/18

Leader: Anthropic · Yes

· 2026-01-15

Audio in/out

11/18

Leader: OpenAI · Partial

· 2026-01-15

Open weights1st

Yes

1/18

· 2026-01-15

OpenAI-compat API1st

Yes

1/18

inference-docs.cerebras.ai · 2026-01-15

Performance2

TTFT p50

200ms

1/2

· 2026-01-15

Output tok/s

2,000 tok/s

1/2

· 2026-01-15

Pricing4

$/M input

$0.35/M tok

4/13

Leader: Groq · $0.05/M tok

inference-docs.cerebras.ai · 2026-11-07

$/M output

$0.75/M tok

4/13

Leader: Fireworks AI · $0.05/M tok

inference-docs.cerebras.ai · 2026-11-07

$/M cache read

$0.35/M tok

9/12

Leader: DeepSeek · $0.0028/M tok

inference-docs.cerebras.ai · 2026-11-07

Batch discount

—

Agentic6

Tool use

Partial

13/18

Leader: Anthropic · Yes

· 2026-01-15

Schema output

Partial

11/18

Leader: Anthropic · Yes

· 2026-01-15

Effort control

12/18

Leader: Anthropic · Yes

· 2026-01-15

Computer use

5/18

Leader: Anthropic · Yes

· 2026-01-15

MCP support

6/18

Leader: Anthropic · Yes

· 2026-01-15

Cache TTL

Automatic, no code changes; guaranteed TTL of 5 minutes, may persist up to 1 hour depending on load

inference-docs.cerebras.ai · 2026-11-07

Governance & continuity6

Continuity policy

Same fragility as Groq, from the same cause: a small curated menu sized to available wafer-scale capacity, with models added and removed as demand shifts. No published notice window. The models themselves are open-weight, so a removal costs you a re-benchmark and a host change rather than a rewrite.

· 2026-01-15

Notice period

—

Pinnable versions

Partial

14/18

Leader: OpenAI · Yes

· 2026-01-15

Zero retention1st

Yes

1/18

cerebras-inference.help.usepylon.com · 2026-11-07