SILMA AI

What SILMA AI does

SILMA AI (SILMA.AI LLC) is a private Arabic AI technology company founded in 2024 and headquartered in Dover, United States, with a 1–10 person team. It builds purpose-built Arabic language AI — including large language models (SILMA 9B Instruct built on Google Gemma, plus the RAG-optimized SILMA Kashif 2B), an Arabic Matryoshka embedding model, and Arabic text-to-speech systems (SILMA TTS v1 open-source and SILMA TTS v2.0 commercial) — and maintains its own evaluation suite (Arabic Broad Benchmark with 470 human-validated questions across 22 tasks, the SILMA RAGQA Benchmark spanning 17 bilingual datasets, and an Arabic TTS Benchmark) hosted on Hugging Face. The technology targets Modern Standard Arabic (Fusha), Saudi Najdi dialect, and Arabic/English code-switching, with the TTS v2.0 product engineered for sub-300ms TTFT streaming, 8 voices, voice cloning, and on-prem/VPC deployment for GCC data residency. Distribution is open-weight via Hugging Face, GitHub, and Ollama, complemented by a self-serve developer portal (dev.silma.ai) and voice platform (app.silma.ai).

The commercial model blends usage-based API access for TTS v2.0, professional services for custom model development and AI evaluation, premium AI consulting engagements, and a subscription-based advisory offering for startups; go-to-market mixes a developer-first PLG motion with direct enterprise field sales into voice-agent/telephony platforms, GCC enterprise and government buyers, and MENA enterprises requiring sovereign Arabic AI. SILMA has reported product cadence across 2024–2025, the Arabic Broad Leaderboard public ranking, and a single disclosed $100,000 funding round in January 2025; no customer logos, ARR, or revenue figures were disclosed in available materials.

Name

SILMA AI

Legal name

SILMA.AI LLC

Website

https://silma.ai

Company type

Private

Founded year

2024

Operating status

Operating

Headcount range

1–10 employees

Short description

SILMA AI is a private US-domiciled Arabic AI company that builds Arabic LLMs, an Arabic embedding model, and Arabic text-to-speech systems — including the sub-300ms TTFT TTS v2.0 — and distributes them via open-weight releases on Hugging Face, GitHub, and Ollama, plus a usage-based API and premium consulting for MENA enterprises, GCC government, and voice-agent platforms.

Ownership category

akta.pro rank

Product category

Arabic AI Models and Text-to-Speech

NAICS

Software Publishers (5132), Custom Computer Programming Services (541511), Computer Systems Design and Related Services (5415)

SIC

Services-Prepackaged Software (7372), Services-Computer Programming, Data Processing, Etc. (7370), Services-Management Consulting Services (8742)

akta.pro primary industry

Model Hosting, Serving & Inference Platforms (HDAAACAB)

akta.pro secondary industries

Enterprise Foundation Model Integration & APIs (Connectors, Governance, Deployment) (HDAAACAO), Fine-Tuning, Adaptation & Custom Model Training (PEFT/LoRA/RLHF) (HDAAACAC), Generative AI & LLM Solutions Services (RAG, Agents, Copilots) (BPAEAHAG), Evaluation, Benchmarking & Observability (Evals, Monitoring, Tracing) (HDAAACAH)

Keywords

• Arabic language models
• Text-to-speech synthesis
• Retrieval-augmented generation
• Embedding model development
• AI consulting services

Headquarters

HQ city

Dover

HQ country

United States

HQ region

North America

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations

Revenue model

1. AI Consulting Services: Premium consulting offerings for building AI for Arabic including speech models, LLM fine-tuning, and evaluation. Targeted at small businesses, startups, and enterprise teams requiring tailored solutions.
2. API Access (TTS and LLM): Usage-based API access for TTS and LLM models via developer portal. Supports streaming modes including Binary Waveform Stream, SSE Audio Stream, and WebSocket Streaming.
3. Custom Model Development: Custom model development services building models tailored to exact specifications, iterating with client teams until performance metrics are met.
4. AI Team Advisory Support: Subscription-based advisory model for fast-moving startups, helping teams design, build, evaluate, and deploy Arabic TTS and LLMs.

Pricing tiers

Go-to-market motion2 records

Distribution channels6 records

Marketing channels9 records

Core offering

SILMA AI builds, fine-tunes, and distributes Arabic-specialized AI models for enterprise and developer use. Its primary offerings are (a) SILMA TTS models (v1 open-source on F5-TTS and v2.0 commercial API with sub-300ms streaming) for Arabic speech synthesis in MSA, Saudi Najdi, and bilingual code-switching; (b) open-weight Arabic LLMs built on Google Gemma, including the general-purpose SILMA 9B Instruct and the RAG-optimized SILMA Kashif 2B Instruct; and (c) embedding models (SILMA Matryoshka) plus proprietary evaluation benchmarks. The company also sells premium consulting services for custom model development, fine-tuning, evaluation, and ongoing AI team advisory.

Product overview

SILMA AI is an Arabic AI technology company offering a comprehensive portfolio of AI products and services. The core offerings include SILMA TTS Models (Text-to-Speech) for Arabic speech synthesis with versions v1 (open-source) and v2.0 (commercial API with sub-300ms streaming), and SILMA Large Language Models including SILMA 9B Instruct (general purpose, 9B parameters) and SILMA Kashif 2B Instruct (RAG-optimized). Supporting products include SILMA Embedding Matryoshka v0.1 for semantic text tasks, the Arabic AI Voice Platform for web-based TTS access, and the Arabic LLM Broad Leaderboard (ABL) for model benchmarking. The company provides AI Consulting Services for speech models, LLM fine-tuning, and evaluation, while the SILMA Research Community (SRC) fosters collaborative open-source AI development. All products are available via APIs (developer portal), open-weight models on Hugging Face, and local deployment via Ollama.

Differentiator

Problem solved

Functional benefit

Products and services

• SILMA TTS v2.0 (Commercial API) Commercial Arabic Text-to-Speech API (v2.1.0) with sub-300ms TTFT streaming delivered via Binary Waveform Stream (POST), SSE Audio Stream (GET), and WebSocket endpoints, with 8 pre-defined voices (sarah, salma, salwa, saja, sultan, salman, sulaiman, salim), voice cloning via custom_audio_id, style/speed control, and on-prem / VPC deployment for MENA voice agent platforms and telephony providers.
• SILMA TTS v1 (Open Source) Open-source bilingual (Arabic/English) text-to-speech model using F5-TTS diffusion architecture, trained from scratch on public and proprietary data and released under a permissive licence for research and commercial use.
• SILMA Saudi TTS Model Foundational Saudi voice engine targeting voice agent platforms, providing hyper-realistic Saudi Najdi and bilingual Arabic/English speech with sub-300ms TTFT streaming and seamless real-time integration, aimed at GCC enterprise and government deployments.
• SILMA 9B Instruct v1.0 Compact 9-billion-parameter open-weight Arabic language model built on Google Gemma that outperforms larger Arabic LLMs in Arabic tasks, released under an open licence for local hosting and commercial use.
• SILMA Kashif 2B Instruct v1.0 Open-weight 2B-parameter LLM built on Google Gemma, optimised for Retrieval-Augmented Generation tasks with up to 12k context length, capable of context-based Arabic and English question answering and entity extraction with top-tier performance within the 3-9B parameter range.
• SILMA Embedding Matryoshka v0.1 Fully-open bilingual (Arabic/English) embedding model using a Matryoshka multi-scale architecture to convert Arabic text into embedding vectors with adjustable dimensions (2 to 768), supporting semantic search, text similarity, and retrieval tasks.
• Arabic AI Voice Platform Self-serve web-based platform providing natural Arabic speech synthesis in Modern Standard Arabic and regional dialects, supporting text-to-speech generation, audiobook creation, and API key management for TTS models.
• AI Consulting Services Premium AI consulting engagements covering Arabic speech-model development, LLM fine-tuning, and model evaluation, complemented by an AI Solution Consultancy Proof-of-Concept and a subscription-based AI Team Advisory for fast-moving startups; offered to small businesses, startups, and enterprise teams.

Quantifiable outcome

• 2-6x lower API costs for TTS compared to alternatives
• +4 more outcomes

Segments4 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

Yes

API docs

API detail

Core technology

AI maturity

App detail

Integration3 records

AI capability5 records

Feature7 records

Scale indicators9 records

Recent moves6 records

Expansion highlights6 records

Direct peers

• ElevenLabs: ElevenLabs is a leading multilingual AI voice/TTS provider with Arabic language support and voice cloning. It is SILMA's primary Western TTS competitor, particularly for voice agent platforms targeting MENA.
• Falcon (Technology Innovation Institute, UAE): Falcon LLM (TII) is an open-weight multilingual LLM with strong Arabic support, hosted via Hugging Face. Like SILMA, it pursues an open-weight + API strategy aimed at Arabic and broader MENA enterprise adoption.
• ALLaM (SDAIA / HUMAIN): ALLaM is the Arabic LLM developed by the Saudi Data and AI Authority (SDAIA) / HUMAIN, targeting sovereign Saudi and pan-Arabic enterprise deployment. It competes head-on with SILMA for Saudi government, telecom, and Arabic enterprise workloads.
• Play.ht: Play.ht is a multilingual AI text-to-speech platform with Arabic voice cloning and API access. It is a direct competitor to SILMA TTS for podcasting, audiobooks, and conversational AI use cases.
• Murf AI: Murf AI offers multilingual AI voice generation with enterprise TTS APIs across many languages including Arabic. It competes with SILMA TTS v2.0 for voice agent, IVR, and conversational AI deployments.
• AI21 Labs: AI21 Labs builds the Jamba family of multilingual foundation models with strong Arabic support and sells enterprise-grade APIs. They target Arabic and GCC enterprise buyers with comparable foundation-model + enterprise GTM motion.
• ACEGPT (BAAI): ACEGPT is an open-source Arabic-centric LLM developed by BAAI/FlagOpen, with public benchmarks and chat capabilities. It competes directly with SILMA 9B and Kashif for open-weight Arabic LLM adoption.
• Inception (Jais): Inception, a G42 company, develops the Jais family of Arabic LLMs (13B/30B/70B) trained natively on Arabic and English. It is SILMA AI's most direct competitor in the foundation Arabic LLM category, competing for the same sovereign GCC enterprise and developer mindshare.

Broad incumbents

• Cohere: Cohere provides enterprise multilingual LLMs (including Arabic) through a managed API and on-prem offerings. It is a broader enterprise AI incumbent competing with SILMA on Arabic enterprise workloads while serving many other languages.
• Amazon Polly: Amazon Polly is a hyperscaler TTS service with Arabic support, deeply integrated into AWS cloud. As a broad incumbent, it competes with SILMA TTS on enterprise and AWS-hosted Arabic voice deployments.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks5 records

Key highlights6 records

Customer concentration

• Xollama.com/silma_ai
• GitHubgithub.com/SILMA-AI
• LinkedInhuggingface.co/silma-ai

Revenue estimate

Valuation estimate

Number of profiles

Profiles1 record

Funding overview

Funding rounds1 record

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about SILMA AI