Model catalog

Every model, one endpoint, one invoice.

49 models from 15 providers, served from EU infrastructure. Prices are per 1M tokens — the same number you see on your invoice, with no regional surcharge.

Get an API key Ask about a model

Models

text, vision, reasoning

Providers

one contract, one key

71ms

Median TTFB

measured in the EU

$0.05

Cheapest input

per 1M tokens

Catalog

Find the model that fits the job.

49 of 49 models

Amazon: Nova Lite 1.0

Amazon · function-calling · vision

Context

300K

Input

$0.06

Output

$0.24

TTFB

87ms

Amazon: Nova Pro 1.0

Amazon · function-calling · vision

Context

300K

Input

$0.80

Output

$3.20

TTFB

61ms

Anthropic: Claude Haiku 5.5

Anthropic · function-calling · vision

Context

1.0M

Input

$0.10

Output

$0.50

TTFB

69ms

Anthropic: Claude Opus 4.8

Anthropic · function-calling · reasoning

Context

1.0M

Input

$5.00

Output

$25.00

TTFB

89ms

Anthropic: Claude Opus 5.5

Anthropic · code · function-calling

Context

1.0M

Input

$4.00

Output

$20.00

TTFB

62ms

Anthropic: Claude Sonnet 5.5

Anthropic · function-calling · vision

Context

1.0M

Input

$2.00

Output

$10.00

TTFB

98ms

Cohere: Command A

Cohere · chat

Context

256K

Input

$2.50

Output

$10.00

TTFB

75ms

Cohere: Command R+ (08-2024)

Cohere · function-calling

Context

128K

Input

$2.50

Output

$10.00

TTFB

73ms

DeepSeek: DeepSeek V3

Deepseek · function-calling

Context

128K

Input

$0.26

Output

$1.03

TTFB

110ms

DeepSeek: DeepSeek V3.2

Deepseek · function-calling · reasoning

Context

131K

Input

$0.28

Output

$0.42

TTFB

55ms

DeepSeek: DeepSeek V4 Flash 0423

Deepseek · function-calling

Context

1.0M

Input

$0.05

Output

$0.45

TTFB

104ms

DeepSeek: DeepSeek V4 Pro 0423

Deepseek · function-calling · reasoning

Context

1.0M

Input

$0.96

Output

$1.91

TTFB

45ms

DeepSeek: R1 0528

Deepseek · function-calling · reasoning

Context

164K

Input

$0.50

Output

$2.15

TTFB

94ms

Google: Gemini 2.5 Flash

Google · function-calling · reasoning

Context

1.0M

Input

$0.30

Output

$2.50

TTFB

39ms

Google: Gemini 2.5 Pro

Google · function-calling · reasoning

Context

1.0M

Input

$1.25

Output

$10.00

TTFB

69ms

Google: Gemini 3.5 Flash

Google · function-calling · reasoning

Context

1.0M

Input

$1.50

Output

$9.00

TTFB

53ms

Google: Gemma 2 27B

Google · chat

Context

Input

$0.65

Output

$0.65

TTFB

95ms

Google: Gemma 3 27B

Google · function-calling · reasoning

Context

131K

Input

$0.08

Output

$0.35

TTFB

72ms

Meta: Llama 3.1 70B Instruct

Meta Llama · function-calling

Context

131K

Input

$0.40

Output

$0.40

TTFB

80ms

Meta: Llama 3.1 8B Instruct

Meta Llama · function-calling

Context

131K

Input

$0.05

Output

$0.08

TTFB

58ms

Meta: Llama 3.2 3B Instruct

Meta Llama · reasoning

Context

131K

Input

$0.05

Output

$0.33

TTFB

91ms

Meta: Llama 3.3 70B Instruct

Meta Llama · function-calling

Context

131K

Input

$0.10

Output

$0.32

TTFB

69ms

Meta: Llama 4 Maverick

Meta Llama · function-calling · vision

Context

128K

Input

$0.19

Output

$0.65

TTFB

113ms

Meta: Llama 4 Scout

Meta Llama · function-calling · vision

Context

328K

Input

$0.10

Output

$0.30

TTFB

75ms

MiniMax: MiniMax-01

Minimax · vision

Context

1.0M

Input

$0.20

Output

$1.10

TTFB

50ms

Mistral Large 2407

Mistralai · code · function-calling

Context

131K

Input

$2.00

Output

$6.00

TTFB

47ms

Mistral: Codestral 2508

Mistralai · code · function-calling

Context

256K

Input

$0.30

Output

$0.90

TTFB

38ms

Mistral: Mistral Large 3 2512

Mistralai · function-calling · vision

Context

262K

Input

$0.50

Output

$1.50

TTFB

92ms

Mistral: Mistral Medium 3.1

Mistralai · function-calling · vision

Context

131K

Input

$0.40

Output

$2.00

TTFB

50ms

Mistral: Mistral Small 4

Mistralai · function-calling · reasoning

Context

262K

Input

$0.15

Output

$0.60

TTFB

70ms

Mistral: Mixtral 8x22B Instruct

Mistralai · function-calling

Context

66K

Input

$2.00

Output

$6.00

TTFB

45ms

MoonshotAI: Kimi K3

Moonshotai · function-calling · reasoning

Context

1.0M

Input

$0.36

Output

$11.90

TTFB

39ms

MoonshotAI: Kimi K3.1 Pro

Moonshotai · function-calling · reasoning

Context

1.0M

Input

$2.00

Output

$10.00

TTFB

74ms

NVIDIA: Nemotron 3 Ultra

Nvidia · function-calling · reasoning

Context

203K

Input

$0.60

Output

$2.40

TTFB

94ms

OpenAI: GPT-4 Turbo

Openai · function-calling · vision

Context

128K

Input

$10.00

Output

$30.00

TTFB

68ms

OpenAI: GPT-4o

Openai · function-calling · vision

Context

128K

Input

$2.50

Output

$10.00

TTFB

62ms

OpenAI: GPT-4o-mini

Openai · function-calling · vision

Context

128K

Input

$0.15

Output

$0.60

TTFB

83ms

Perplexity: Sonar

Perplexity · vision

Context

127K

Input

$1.00

Output

$1.00

TTFB

113ms

Qwen2.5 72B Instruct

Qwen · function-calling

Context

33K

Input

$0.36

Output

$0.40

TTFB

104ms

Qwen2.5 Coder 32B Instruct

Qwen · code · reasoning

Context

33K

Input

$0.66

Output

$1.00

TTFB

60ms

Qwen: Qwen2.5 VL 72B Instruct

Qwen · vision

Context

128K

Input

$0.80

Output

$1.00

TTFB

106ms

Qwen: Qwen3 235B A22B Instruct 2507

Qwen · function-calling

Context

262K

Input

$0.09

Output

$0.55

TTFB

74ms

Qwen: Qwen3 Coder 480B A35B

Qwen · code · function-calling

Context

262K

Input

$0.30

Output

$1.00

TTFB

85ms

Qwen: Qwen3.7 Max

Qwen · function-calling

Context

1.0M

Input

$1.48

Output

$4.42

TTFB

60ms

Qwen: Qwen3.7 Plus

Qwen · function-calling · vision

Context

1.0M

Input

$0.32

Output

$1.28

TTFB

71ms

Qwen: Qwen3.8 Flash

Qwen · code · function-calling

Context

1.0M

Input

$0.15

Output

$0.47

TTFB

70ms

Qwen: Qwen3.8 Max Prime

Qwen · function-calling · vision

Context

1.0M

Input

$4.00

Output

$12.00

TTFB

103ms

SpaceXAI: Grok 4.5

X Ai · function-calling · vision

Context

500K

Input

$2.00

Output

$6.00

TTFB

38ms

Z.ai: GLM 5.3 Prime

Z Ai · function-calling

Context

1.0M

Input

$2.80

Output

$8.80

TTFB

42ms

Starting points

If you don't want to compare 49 rows

Lowest cost

Meta: Llama 3.1 8B Instruct

Meta Llama · 131K ctx

At $0.05 in and $0.08 out per 1M tokens, this is the cheapest way to run high-volume classification, extraction, and routing work.

Longest context

DeepSeek: DeepSeek V4 Flash 0423

Deepseek · 1.0M ctx

1.0M tokens of context — enough to hold a mid-sized codebase, a contract set, or a full support history in one request.

Fastest first token

Mistral: Codestral 2508

Mistralai · 256K ctx

38 ms to first token, measured from EU regions. The right default for interactive interfaces where perceived speed matters more than depth.

Integration

Three lines to switch, no rewrite

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.itciphtech.com/v1",
    api_key="cph_live_...",
)

r = client.chat.completions.create(
    model="deepseek/deepseek-v4-flash",
    messages=[{"role": "user", "content": "Summarise this invoice."}],
)

cURL

curl https://api.itciphtech.com/v1/chat/completions \
  -H "Authorization: Bearer $CIPHER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4-flash",
    "messages": [{"role":"user","content":"Summarise this invoice."}]
  }'

Streaming, tool calling, JSON mode, and embeddings follow the same schema. Switching model is a string change; switching back is another one.

Comparison

What changes when you buy through us

	Direct with each provider	CIPHER
Contracts	One per vendor	One
Invoicing	Multiple currencies and entities	One EUR invoice
Data residency	Provider-dependent regions	EU by default
API shape	Mostly compatible, not identical	Single OpenAI-compatible schema
Rate limits	Per-vendor, renegotiated	Pooled across the catalog
Support	Ticket per vendor	One team, one thread
Failover	You build it	Per-model fallback

Provider terms still apply to the underlying model weights and licences; the commercial relationship for access, billing, and support is with CIPHER TECH B.V.

FAQ

Questions we get before the first key

Do I need a new SDK to use these models?

No. The endpoint is OpenAI-compatible, so the official OpenAI SDKs work with a different base URL and key. Anything that speaks the chat-completions format — LangChain, LlamaIndex, Vercel AI SDK — works unchanged.

Where exactly do requests run?

In European data centres operated for CIPHER. Prompts and completions are processed in the EU and are not routed to a provider region outside it. Inference logs are kept according to the retention period in your contract.

Is there a minimum commit or a monthly platform fee?

No platform fee and no minimum commit on the pay-as-you-go catalog. You are billed per token at the rates shown. Dedicated capacity and the gateway have their own plans, priced separately.

What happens when a provider has an outage?

Each model has a defined fallback in the same capability class. If you use the gateway, failover is automatic and recorded in the request trace; without it, you can pin a fallback model yourself.

Can you host a model that is not in this list?

Yes — open-weight checkpoints and your own fine-tunes can be deployed on dedicated GPUs, including in an air-gapped configuration. That is the private hosting product, not the shared catalog.

How is cached input billed?

Where a model supports prompt caching, cached input tokens are billed at the discounted rate shown on the model page. The cache is scoped per API key and expires automatically.

Need a model that isn't listed?

We host custom and fine-tuned weights on dedicated GPUs, inside your own network boundary if required.

Talk to us