# reka-flash-3.1

[Canonical model page](https://headlessthrones.com/models/reka-reka-flash-3-1) · [Model library](https://headlessthrones.com/models)

Public catalog reference from Headless Thrones. Exact catalog ID: `reka-reka-flash-3-1`. Catalog source-check date: 2026-10-23.

This brief contains the public catalog answers and any catalog guide, not private drafts, historical editorial revisions or independently measured rankings. The canonical HTML page may also display separately published editorial.

Current catalog answers

## What to know about reka-flash-3.1

AI-compiled from linked model documentation. This is a catalog reference, not an independent performance review. Catalog checked October 23, 2026.

[Exact model](https://headlessthrones.com/models/reka-reka-flash-3-1#model-answer-identity) · [Access](https://headlessthrones.com/models/reka-reka-flash-3-1#model-answer-access) · [Inputs & outputs](https://headlessthrones.com/models/reka-reka-flash-3-1#model-answer-modalities) · [License](https://headlessthrones.com/models/reka-reka-flash-3-1#model-answer-license) · [Model details](https://headlessthrones.com/models/reka-reka-flash-3-1#model-answer-research) · [Weight details](https://headlessthrones.com/models/reka-reka-flash-3-1#model-answer-repository)

### Which reka-flash-3.1 does this page describe?

reka-flash-3.1 is a model record from **Reka** in the reka family. [Source](https://huggingface.co/RekaAI/reka-flash-3.1)

The recorded model identifier or checkpoint is `reka-flash-3.1`. Use that exact identifier when checking documentation or a provider listing. [Source](https://huggingface.co/RekaAI/reka-flash-3.1)

### How can you access reka-flash-3.1?

The catalog documents **Open weights** access  for reka-flash-3.1. [Source](https://huggingface.co/RekaAI/reka-flash-3.1)

As of the catalog check, this version was listed without confirmed general availability. [Source](https://huggingface.co/RekaAI/reka-flash-3.1)

Open weights means the model weights are distributed for use under their stated terms. It does not establish a hosted API, an open-source license, or a hardware requirement.

### What can reka-flash-3.1 receive and generate?

Documented inputs: **Text**. [Source](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

Documented outputs: **Text**. [Source](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

Input support and output generation are separate capabilities. Reading an image does not establish image generation; a text output listing does not establish speech output. File sizes, formats and media limits depend on the documented implementation.

### What license is recorded for reka-flash-3.1?

The recorded license label is **apache-2.0**. [Source](https://huggingface.co/api/models/RekaAI/reka-flash-3.1)

Read the linked license and associated use terms for this exact version before redistribution, commercial deployment or fine-tuning. A label alone does not establish every permission, restriction or obligation, and it does not determine the terms of a separate hosted service.

### What distinguishes reka-flash-3.1?

These details come from the linked publisher material for this model. Reported capabilities and training descriptions are publisher statements, not independent test results.

**Model specialization**: Coding-focused update that can serve as a starting point for agentic fine-tuning. [Source](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

**Training procedure**: Post-training combines supervised fine-tuning on synthetic and public data with RLOO reinforcement learning using verifiable rewards. [Source](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

**Tokenizer**: Uses cl100k\_base without additional special tokens. [Source](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

**Language scope**: English is the intended supported language; the publisher does not guarantee other-language capability. [Source](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

**Conversation handling**: Stop generation at <sep> or the end-of-text token; the card recommends removing earlier assistant reasoning traces before the next conversation turn. [Source](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

### What is documented about the reka-flash-3.1 checkpoint?

This evidence refers to `RekaAI/reka-flash-3.1` at recorded revision `dcf8d6ee3219ca177b519cab47777d42864b54bd`, checked October 23, 2026. [Repository metadata](https://huggingface.co/api/models/RekaAI/reka-flash-3.1)

Safetensors metadata records **20,905,482,240 stored parameters**. This is a stored-weight count, not necessarily the number activated per token, a download size or a GPU-memory requirement. [Repository metadata](https://huggingface.co/api/models/RekaAI/reka-flash-3.1)

Recorded architecture class: `LlamaForCausalLM`. [Repository metadata](https://huggingface.co/api/models/RekaAI/reka-flash-3.1)

Repository access gate: `false`. This records the repository's gate setting when checked; licensing and use terms still apply. [Repository metadata](https://huggingface.co/api/models/RekaAI/reka-flash-3.1)

#### What the configuration file says

The pinned `config.json` contains these values. [Pinned configuration](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/config.json)

`max_position_embeddings`: 98,304

`torch_dtype`: bfloat16

`num_attention_heads`: 64

`num_key_value_heads`: 8

`num_hidden_layers`: 44

`hidden_size`: 6,144

Configuration values describe the checkpoint and its implementation. In particular, max\_position\_embeddings, sliding\_window and rope\_scaling do not independently establish an API context limit, tested usable length or hardware requirement. Runtime, quantization, cache, batch size and input length affect deployment behavior.

Read the retrieved model card for setup and qualifications. [Model card](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md)

**Catalog qualifications:** Official publisher inventory entry. Missing fields have not been verified. Repository timestamps are not release dates. [Source 1](https://huggingface.co/api/models/RekaAI/reka-flash-3.1) · [Source 2](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/README.md) · [Source 3](https://huggingface.co/RekaAI/reka-flash-3.1/resolve/dcf8d6ee3219ca177b519cab47777d42864b54bd/config.json)

**Evidence still missing here:** context or input capacity, knowledge and training-data cutoffs, and tested hardware requirements. Missing documentation is not evidence that a feature is unavailable.

Headless Thrones has not independently benchmarked this model for this reference. Evaluate it on representative work before choosing it; these facts do not rank its accuracy, reliability, speed or suitability against another model.

Keep the exact version in view

## Other reka models from Reka.

These are separate records in the same publisher family. A shared family name does not establish equal capabilities or interchangeable licenses.

### [reka-edge-2603](https://headlessthrones.com/models/reka-reka-edge-2603)

API · Open weights · 16,384 context tokens

### [reka-flash-3](https://headlessthrones.com/models/reka-reka-flash-3)

API · Open weights · 64,000 context tokens