Flam

Flam is bringing AI interactive content format to ADWEEK 2026. We’ll be in Atlanta, September 15–17, connecting with the marketers and brand leaders shaping what’s next. Swipe through to meet the Flam team and get a glimpse of the conversations we’ll be having around AI, interactive content format, engagement, media, and the future of digital experiences. Going to Brandweek? Book time with the Flam team: Kumar Shah, Ryan Clough, Monique C. & Aditya Kansal https://lnkd.in/d4y5zKXe #Brandweek #adweek #Atlanta #Flam

Everyone publishes benchmark numbers. Almost nobody publishes the harness. We’re publishing both. Meet Falcon -our Mixture-of-Experts model with 26B total parameters and ~4B active per token. And instead of asking you to trust the numbers, we’re giving you everything you need to reproduce them. All performance below was measured on a single RTX PRO 6000, under identical load: 82.0 MMLU-Pro Within 3 points of DeepSeek-V3.1’s published 84.8 with only ~4B parameters active per token. 2,029 decode tokens/sec. 30ms TTFT. The next best model in our panel: 1,699 tok/s. Llama-3.3-70B: 152 tok/s. $0.42 / 1M output tokens ~13× cheaper than Llama-3.3-70B. ~5× cheaper than Gemma-3-27B. 83.2 MILU composite across 10 Indian languages Best in the panel with Falcon ahead of the closest competitor on every language tested. 506 end-to-end RAG test cases → 100% gold-chunk Recall@10 → 92.9% factual accuracy → 7.9% hallucination rate And there’s an important detail behind those RAG numbers: Every test case includes a verbatim evidence span from the source chunk, verified using exact string matching. The model generating the answer never gets to grade itself. Because a benchmark you can’t reproduce is just a claim. So we open-sourced the evidence. The full Falcon benchmark suite, methodology, and results are public. 👉 Tap the link: https://lnkd.in/dRkrhtuu #Falcon #FlamAI #LLM #MixtureOfExperts #AIBenchmarks #MMLU #OpenBenchmarks

We are hiring in SF & NY: AI Engineers. Product Managers. Enterprise Sales. Chief of Staff. https://lnkd.in/dEQSkV7z #hiring #US #SF #NY

Nigeria, that was fun. Flam in partnership with SmartSense Technologies, brought our AI interactive content platform to GITEX NIGERIA - The Largest Tech & Startup Event in West Africa 2026, putting it in front of innovators, creators and tech leaders from across the region. This is what Flam looks like in a room full of people experiencing it for the first time. Oluseyi Adebayo-Olubi #Flam #GitexNigeria #Gitex2026 #gitex

We’re excited to have Ryan Clough join us as our Enterprise Account Director, bringing his expertise to our growing team in San Francisco. Welcome to the team, Ryan! #Flam #USteam #SF

The next big content format is being built. Want to help define it? We’re hiring a Senior Product Manager in San Francisco to build the future of AI interactive content. High ownership. Close to the founders. Building from 0→1. If that sounds like you, let’s talk. Apply here : https://lnkd.in/enJqgKDz #hiring #SanFrancisco #US #hiringproductmanager #flam

What if AI didn’t just answer back, but knew what the answer should look like? Ask Falcon for your sales numbers. You get a chart. Ask it to make sense of a diagram. You get an artifact. Need something done? It gets it done. Falcon knows the difference in between something you need to read, see or do and responds accordingly. Curious? Early access is open. Tap the link. https://lnkd.in/dCxs7_7W #flam #falcon #artifacts #interactivecontent #ConversationalAI

Indic languages are the toughest to crack for a multilingual LLM. India speaks 22 languages. AI breaks on almost all of them. The best English models struggle with Hindi and barely understand Odia, Assamese or Sanskrit. The specialists that do speak them well are too slow for real-time use. Rupam on how Falcon got native Indic fluency without losing anything the model already knew. Curious? Early access is open to Indian enterprises. Tap the link. https://lnkd.in/dCxs7_7W #Flam #Falcon #MultilingualAI #Indic #LLM

Most models translate first and think second. That's why they drift out of your language mid-sentence and break the formats your product depends on. Rupam M. on how Falcon gets to first token in 29ms on average and why that number doesn't move when the language does. Early access is open. https://lnkd.in/dkNSnzfC