25 signals
HOlO V1 IS LIVEone ranked AI digest a day, scored in publicREAD HOW IT WORKS →HOlO V2 STARTSyour X account, your signals, every day

Model registry / Cohere / Tiny / Tiny Aya Fire

Tiny Aya Fire

Compact multilingual model specialized for South Asian languages

Line

Tiny

Weights

Open

Released

2026-02-17

Context

8K

Input

not stated

Output

not stated

Max output

8K

Coverage

not covered yet

Overview

Tiny Aya Fire is a compact multilingual model from Cohere, specialized for South Asian languages. It is part of the Tiny line, following Tiny Aya Earth. It has a context window of 8000 tokens and a maximum output of 8000 tokens. It accepts text and returns text, supports tool calling, and does not have a reasoning mode. The weights are open. It is served by 1 host. Pricing is not stated.

Specs

Released2026-02-17
LineCohere · Tiny
WeightsOpen
Context8K tokens
Max output8K tokens
Inputtext
Outputtext
ReasoningNo
Tool callingYes
Knowledge cutoffnot stated
API idtiny-aya-fire

Pricing

Inputnot stated
Cached inputnot stated
Outputnot stated

First version of its line in the registry.

Strengths

  • Compact model for South Asian languages
  • 8000-token context and output window
  • Supports tool calling
  • Open weights
  • Text-to-text mode

Best for

  • Reach for it for multilingual tasks in South Asian languages
  • Reach for it for text generation with tool calling
  • Reach for it for compact deployment with open weights

How to access

ProviderModel id
Coheretiny-aya-fire

1 host serve this model at the price above · the maker's documentation

Tiny: every version

VersionReleasedContextInput
Tiny Aya Fire2026-02-178K
Tiny Aya Global2026-02-178K
Tiny Aya Water2026-02-178K

The registry keeps the six most recent versions of a line. Older ones stay in the file and can be surfaced without a rebuild.

FAQ

What is the context window?
The context window is 8000 tokens, and the maximum output is also 8000 tokens.
Does it support tool calling?
Yes, it supports tool calling. It does not have a reasoning mode.
Are the weights open?
Yes, the weights are open. It is served by 1 host.
Open weights8K context