Meta 7/13 Llama 4 Behemoth refresh: 1M long-context + code-weight under Llama 4 Community License

2 sources

Meta on 2026-07-13 publishes a Llama 4 Behemoth refresh emphasizing long-context (1M + token-efficiency) and a code-weight specialist. Self-reported figures include MMLU-Pro 83.4, GPQA-Diamond 75.6, SWE-Bench Verified 69.7, HumanEval+ 91.2, Terminal-Bench 2.1 75.9. License: Llama 4 Community License with the commercial-use restrictions addendum.

Why now

It's the first Meta anchor in the current cluster — combines with the OpenAI gpt-oss refresh to give 'US-lab open-weight dual.' Creators who cover the Meta long-context + coding beat can ship on this without waiting.

Why it is worth publishing

Best for long-context creators + open-weight-distribution creators. The Llama 4 Community License addendum is the legal hook; the 1M context is the deployment hook.

Evidence basis

First Meta anchor of the cluster + commercial-use addendum on top of Llama 4 Community License + 1M context. Triple heat; topPick-eligible but already allocated (gpt-oss + Sonnet 4.5 share the badge slots).

“Meta just shipped a fresh Llama 4 — 1M context, code-weight specialist, and a commercial-use addendum you should read before quoting the headline.”

Angle

Frame Llama 4 Behemoth refresh as Meta's long-context + code-weight specialist move — not as 'Llama 5.' Lead with the architecture framing (1M context + code-weight specialist), then the license framing (Llama 4 Community License + commercial-use restrictions addendum).

Format

Carousel + a single-card distillation

Demo idea

Carousel card 1: 'Llama 4 Behemoth refresh — what changed'; card 2: '1M context + code-weight specialist'; card 3: 'HumanEval+ 91.2 / SWE-Bench Verified 69.7 / Terminal-Bench 2.1 75.9 — vendor-supplied, not reproduced'; card 4: 'Llama 4 Community License + commercial-use restrictions addendum — quote addendum language, don't paraphrase'; card 5: 'vs Llama 4 prior: refresh, not a base-model swap — frame accordingly'. Pairs cleanly with the OpenAI gpt-oss card to make a 'US open-weight dual'.

Platform notes

Refresh vs base-model swap (medium risk): do not call Llama 4 Behemoth refresh 'Llama 5' or 'a new base model.' When viewers ask 'is this Llama 4 better than Llama 4 prior at coding,' pin the link to the Meta blog and quote only the deltas that are stated; do not generalize from the vendor-stated deltas. Commercial-use restrictions boundary was not extracted in the captured summary.

Usable claims

Meta on 2026-07-13 published a Llama 4 Behemoth refresh emphasizing long-context (1M context + token-efficiency) and a code-weight specialist; self-reported figures include MMLU-Pro 83.4, GPQA-Diamond 75.6, LiveCodeBench v6 73.8, SWE-Bench Verified 69.7, HumanEval+ 91.2, Terminal-Bench 2.1 75.9; license is the Llama 4 Community License with the commercial-use restrictions addendum.

Evidence pipeline

From the news

Meta 7/13 Llama 4 Behemoth refresh: 1M context + code-weight specialist under Llama 4 Community License

Breakdown

Walks the Llama 4 Behemoth refresh's long-context + code-weight framing (1M + token-efficiency / code-weight specialist), benchmark headlines (HumanEval+ 91.2 / SWE-Bench Verified 69.7 / Terminal-Bench 2.1 75.9), and the Llama 4 Community License + commercial-use restrictions addendum — paired with one explicit 'refresh, not a base-model swap' qualifier. Addendum scope was not extracted.

Sources

Risks

Pin the link to each vendor page and the companion HF model card; quote only what those pages state; do not paraphrase self-reported benchmark numbers as third-party validation; for the Doubao topic specifically, pin the HF mirror card as a researcher-preview only and do not paraphrase the gating-to-API-only boundary.
Pin the link to each license text; quote only what the page states; do not paraphrase the baseline license as the sole operating condition; flag addenda by name without stating their scope; for Google Gemini 3.5 Pro, note that license lives with the API contract rather than a downloadable-weights license.

Demo ideas

Pair with OpenAI gpt-oss refresh — show as 'US-lab open-weight dual, both Apache-2.0-or-Llama-4-Community-License-with-addenda,' with one column per vendor and one for the addendum name.
For long-form Meta coverage, build a 'Meta Llama 4 mid-2026 refresh timeline' — flag the addendum language carefully and avoid generalizing vendor-stated deltas.
Use as the cluster's 'first Meta anchor' hook: 'today's six-topic cluster adds Meta to the corpus for the first time — Llama 4 Behemoth refresh, with commercial-use restrictions addendum.'