Is it winning?
Yes
Assessed capability across ten dimensions · 65.0
Loading this page
Assembling this page. Live sources are being read now.
Assembling this page. Live sources are being read now.
Wednesday, 7 October 2026
Llama owner with deep open-weights distribution and Meta-AI consumer product surface.
Yes
Assessed capability across ten dimensions · 65.0
Yes
Customer, developer and employee reputation · 78.0
Yes
What the company discloses about its finances · 85.0
Analyst interpretationDecisive for open-weights strategies; less of a packaged enterprise platform play.
| Capability | maturityScore | status | evidenceGrade | lastVerified |
|---|---|---|---|---|
| agents | 50.9 | verified | E3 | 2026-07-30 |
| deployment_options | 81 | tested | E3 | 2026-05-07 |
| portability | 60 | verified | E4 | 2026-08-07 |
| enterprise_assistant | 49 | verified | E3 | 2026-07-30 |
| governance | 86 | verified | E4 | 2026-05-07 |
| security | 52.9 | verified | E5 | 2026-07-30 |
| rag | 76 | tested | E3 | 2026-05-07 |
| models | 59.4 | verified | E4 | 2026-08-07 |
| cost_controls | 76 | tested | E3 | 2026-05-07 |
| integrations | 58.5 | verified | E3 | 2026-08-07 |
Not covered in the capability matrix
The AI Enterprise capability matrix does not include Meta. No capability scores are shown rather than invented.
Meta is one of NVIDIA's largest customers, public guidance of 350K+ H100 GPUs by end of 2024 for Llama training and Reality Labs.
Llama 3.x hosted on Groq LPU inference, flagship open-model partnership announced at Meta Connect.
Together AI is a flagship Llama inference + fine-tuning host, full Llama 3.x family available.
Fireworks AI is a Meta launch partner for Llama enterprise inference + fine-tuning.
Meta is a flagship AMD Instinct MI300X customer, announced large-scale Llama inference deployment on MI300.
Llama 3.x family hosted on Bedrock; Meta retains ownership.
Meta's 350K+ H100 fleet for Llama training is fabricated by TSMC; MTIA accelerators are also on TSMC N5.
Llama 3.x served via Cerebras Inference cloud at industry-leading tokens/sec, open-model inference channel.
Llama models hosted on Azure AI Foundry, Meta remains the original owner.
OCI Generative AI hosts Llama models; Meta retains ownership.
Meta has used CoreWeave for burst GPU capacity beyond its in-house clusters; relationship not formally disclosed.
| Model | modelFamily | modelCategory | availabilityStage | ownershipType | dataStatus | Source |
|---|---|---|---|---|---|---|
| Llama 4 | Llama | multimodal | ga | first_party | seed · E2 | www.llama.com |
| Llama 4 Scout | Llama | llm_text | ga | first_party | seed · E2 | www.llama.com |
| Llama (hosted on Bedrock) hosted by Amazon | Llama | llm_text | ga | hosted_third_party | seed · E2 | docs.aws.amazon.com |
| Llama (hosted on OCI) hosted by Oracle | Llama | llm_text | ga | hosted_third_party | seed · E2 | docs.oracle.com |
github.com/meta-llama · reddit.com/r/LocalLLaMA
glassdoor.com/Overview/Working-at-Meta · blind
g2.com/products/llama
No curated sources listed
The sourcing manifest holds no curated evidence URLs for Meta.