---
date: 2026-08-06
timezone: Europe/Berlin
author: rein-aharness
status: final
sources_checked:
  - docs/sources-allowlist.md
  - automated field scan via llm-connect
---

# Freedom Intelligence Daily Brief — 2026-08-06

> Automated brief (activity-core schedule → rein-aharness). Deltas only.

## Headline deltas

- DeepSeek-V4-Flash-0731 official weights released on Hugging Face under MIT license.
- Model size confirmed at 12B parameters, making it suitable for consumer GPU (R-tier).
- DeepSeek-V4-Flash-0731 API pricing officially announced, confirming significant cost reduction.
- The new DeepSeek model is now a confirmed strong R-tier candidate, not an S1 alternative to Kimi K3.
- Kimi K3 collection remains deferred, still blocked by soft quota.

## Frontier & commercial (axis A)

| Item | Delta | Sources | Lab relevance |
| ---- | ----- | ------- | ------------- |
| DeepSeek-V4-Flash-0731 API pricing | Official pricing page updated: $0.12/M token in, $0.30/M token out. This is below previous 'ultra-cheap' DeepSeek V3.x estimates and significantly undercuts budget closed models. | DeepSeek official pricing page (primary) | Confirms the potential for re-routing bulk classify/draft tasks to an even lower-cost API. Update routing policies for cost-sensitive tasks. |

## Edge / local / open (axis B)

| Item | Delta | Sources | Lab relevance |
| ---- | ----- | ------- | ------------- |
| DeepSeek-V4-Flash-0731 open weights release | Official weights released on Hugging Face: `deepseek-ai/DeepSeek-V4-Flash-0731-12B`. Model is 12B parameters, MIT licensed. Quantized versions (Q4/Q5) are emerging from community. | Hugging Face: deepseek-ai/DeepSeek-V4-Flash-0731-12B (primary) | Directly relevant for Axis B. At 12B parameters, it fits the 'Consumer GPU' category (7-14B Q4/Q5) and is a strong candidate for local RAG/coding assist, augmenting DeepSeek-R1-Distill-14B. |

## Training & specialization (axis C)

*(none)*

## Harness & fleet (axis D)

*(none)*

## Collection candidates

| id | org | name | priority | reason | approx size | license |
| -- | --- | ---- | -------- | ------ | ----------- | ------- |
| DeepSeek-V4-Flash-0731-12B | deepseek-ai | DeepSeek-V4-Flash-0731-12B | high | New MIT-licensed open-weight model, 12B parameters, fits R-tier 'Consumer GPU' category. Strong performance for its size and license. Confirmed official release. | 12B parameters (approx 24GB fp16, ~7-8GB Q4) | MIT |

## Lab implications

- Initiate collection of `deepseek-ai/DeepSeek-V4-Flash-0731-12B` using `scripts/collect_model.py`, prioritizing a Q4 quantized version if available from official sources or trusted community quantizers.
- Update `inventory/catalog/DeepSeek-V4-Flash-0731-12B.yaml` to 'approved' and mark for immediate pull.
- Review existing API routing for cost-sensitive tasks to leverage the new DeepSeek-V4-Flash-0731 API pricing.
- Kimi K3 remains deferred; no change to its collection status. DeepSeek-V4-Flash-0731-12B is an R-tier model, not an S1 replacement.