Co-Founder at @Phonic_Co. Previously @Stanford CS PhD Dropout, @MosaicML, CS @MIT. I tend to be wrong, but the learning process makes it enjoyable. 🇵🇰🇺🇲
San Fransisco Bay Area
Joined October 2009
- Tweets18K
- Following1.1K
- Followers3.5K
- Likes5.2K
Pinned Tweet
Introducing Phonic-1, the most advanced speech-to-speech model for agentic tasks — from @phonic_co, the speech-to-speech model company.
We’ve had a breakout year and raised a $27.5M Series A led by @firstround with participation from @Lux_Capital, @felicis, and Definition.
Moin Nadeem retweeted
10 people! Go @phonic_co 🔥🔥
OpenAI just got beaten at voice by 10 people.
Not a lab with 1,000 researchers. 10.
Phonic-1 took #1 on the speech-to-speech benchmarks, ahead of gpt-live-1. And it's $0.06/min.
The voice agent shortlist just changed.
Thanks for your support @TobennaA! It means a ton.
Good and extremely sharp team. When our previous voice vendor was getting acquired, we needed to find an alternative system asap, @moinnadeem and team moved so fast, and impressed us so much that my cofounders and I did something we rarely do - decided to not just be customers but also asked to write personal checks into the company. 🙏🏾
Moin Nadeem retweeted
Even if I didn't care about deep learning, I'd be happy about this because Moin is a great guy. But this is also a very strong release on its own, let alone accomplished by a team of only ten people. Big up to the @phonic_co!
Introducing Phonic-1, the most advanced speech-to-speech model for agentic tasks — from @phonic_co, the speech-to-speech model company.
We’ve had a breakout year and raised a $27.5M Series A led by @firstround with participation from @Lux_Capital, @felicis, and Definition.
Moin Nadeem retweeted
Pretty insane to see Phonic so good on the agentic performance side while also having such low latency
Phonic's Phonic 1 (High) is the new leading Speech to Speech model, debuting at #1 on the Artificial Analysis Speech to Speech Index at 87.1, while also leading on agentic performance and being competitive on latency
Phonic 1 is @phonic_co 's Speech to Speech model, available through the Phonic API at $0.07 per minute. We evaluated Phonic 1 at its high intelligence level through the Phonic API.
Key takeaways:
➤ Speech to Speech Index: #1 of 22 at 87.1, ahead of Grok Voice 3.0 (86.9) by a very small margin. The Index averages Speech Reasoning (Big Bench Audio), Agentic Performance (Tau Voice), Arena Preference and Task Success Rate
➤ Speech Agent Arena: #8 of 30 on preference (Elo 1012) and #4 on Task Success Rate (92.3%)
➤ Tau Voice: #1 at 79.9%, a clear lead of 5.4 points over the next model
➤ Big Bench Audio: #3 at 98.9%, behind StepAudio 3 Realtime (99.7%) and Qwen Audio 3.0 Realtime Plus (99.2%)
➤ Speed: #4 on Time to First Audio at 0.61s, and the fastest of the top 5 models in the Index
➤ Cost: $4.20 per hour of input audio
Congrats to the @phonic_co team 🚀
See below for more detail ⬇️
- Phonic released a model which outperforms GPT-Live
🚀
Moin Nadeem retweeted
there's a new #1 voice agent on VAmoS Bench 🥇
we benchmarked @phonic_co's Phonic-1 on ultra-realistic voice simulations from @veris_ai
TLDR;
-> 57.5% pass rate (next best: ~45%)
-> $0.20 per call (middle of the pack)
-> strong in street / cafe noise; TV still hard
🧵/
Moin Nadeem retweeted
Phonic's Phonic 1 (High) is the new leading Speech to Speech model, debuting at #1 on the Artificial Analysis Speech to Speech Index at 87.1, while also leading on agentic performance and being competitive on latency
Phonic 1 is @phonic_co 's Speech to Speech model, available through the Phonic API at $0.07 per minute. We evaluated Phonic 1 at its high intelligence level through the Phonic API.
Key takeaways:
➤ Speech to Speech Index: #1 of 22 at 87.1, ahead of Grok Voice 3.0 (86.9) by a very small margin. The Index averages Speech Reasoning (Big Bench Audio), Agentic Performance (Tau Voice), Arena Preference and Task Success Rate
➤ Speech Agent Arena: #8 of 30 on preference (Elo 1012) and #4 on Task Success Rate (92.3%)
➤ Tau Voice: #1 at 79.9%, a clear lead of 5.4 points over the next model
➤ Big Bench Audio: #3 at 98.9%, behind StepAudio 3 Realtime (99.7%) and Qwen Audio 3.0 Realtime Plus (99.2%)
➤ Speed: #4 on Time to First Audio at 0.61s, and the fastest of the top 5 models in the Index
➤ Cost: $4.20 per hour of input audio
Congrats to the @phonic_co team 🚀
See below for more detail ⬇️
Moin Nadeem retweeted
Phonic takes gold on Tau Voice English!
Its the first time that a model not produced by one of the big labs (openai/xai/gemini) has come out on top. Congrats to @NikhilKMurthy, @moinnadeem and the Phonic team, this is huge!
Phonic-1: lnkd.in/p/gfPHZztN
Leaderboard: taubench.com/leaderboard/?be…
Moin Nadeem retweeted
A few years ago, Moin and I were driving down 280 and talking about building something in voice AI. We were excited by the future with models that could truly understand and naturally speak with us, and it felt obvious to me at the time that these models, like us, needed to take in and output audio. It turns out this is an extremely hard thing to build, but I’m so proud of the Phonic team for taking us a step closer in that direction with Phonic-1.
Few lessons I learned:
1. Intelligence cannot be traded off for low latency or naturalness: the most valuable workloads in voice are intelligence bottlenecked (multi-step tool calling, instruction following and long context understanding)
2. Real calls are the ultimate benchmark. Whenever I sit with a customer and listen to their calls, I always find new failure modes and learnings that help us build at the frontier.
3. Reliability is everything. A voice call going silent or dropping is truly an awful experience, and you can’t just retry a request (like coding agents). We had to think about our infra from the ground up for reliability.
We’re so excited to release Phonic-1. It’s a really great model across intelligence, latency and naturalness, and already making a huge impact with our customers. And so much more in store to come!
Introducing Phonic-1, the most advanced speech-to-speech model for agentic tasks — from @phonic_co, the speech-to-speech model company.
We’ve had a breakout year and raised a $27.5M Series A led by @firstround with participation from @Lux_Capital, @felicis, and Definition.
We’re a ten-person team of researchers and engineers from MIT, Stanford, and CMU, including International Olympiad medalists and former founders.
We believe that building at the frontier means working deeply with customers, because the thorniest problems in voice only show up in real calls. Solving them has meant rethinking every layer of the stack: data, pre-training, RL, kernels, and infrastructure.
The result is Phonic-1, a model that doesn’t trade intelligence for naturalness.
Phonic-1 is available today.
Get $250 of free credits if you sign-up by Nov 30. Already live with voice? Customers have switched to Phonic from cascaded stacks on ElevenLabs, Bland, and others. If you switch, three months are on us.
Read more: phonic.ai/blog/phonic-1
Masterclass.
I don't think we should fight. The two companies are complementary. Since you're not doing this AI stuff, I really think Snowflake should consider getting its AI tokens through Databricks. You can get both cheap OSS models as well as frontier (OAI/Ant/Gemini/Grok) models from our platform. It all goes through Unity Gateway which has Smart Routing, Budgets, and Access Control. You can even use Omnigent with any harness, including CoCo. Most of our customers used this to reduce token spend ~30% last few months while token count went up. Genie lets you then track, control, manage that spend, all your financials, marketing, etc. Quick POC?
There's two main axes to make inference more efficient: increase tokens per second, or just use fewer tokens.
Everyone has been chasing the first, but there haven't been enough people chasing the second.
In a compute shortage, @Suhail has the right approach: if you use fewer tokens, you can get away with fewer GPUs. This problem is boundless and I'm excited to see him make a dent in the mission.
1/ Excited to announce our new company Particle
This is our first step in a series of models focused on developers.
With particle-1, we focused on token efficiency and managed to keep parity in performance while using ~45% fewer tokens.
Live now: particle.ai
I don't think an interface is going away, I just think voice is going to be used to drive a UI instead of chat.
Interfaces are not going away. Some people keep saying the future is pure chat. "No UI, just talk to it."
If that were true:
- no meeting would have a whiteboard
- pitch decks wouldn’t exist
- You wouldn’t look at your phone screen
You always need to show things and interact visually: drawing lines, previewing, whatever. Those needs don’t disappear because AI got better.
Gartner: so yeah there's more GPUs than demand folks, I think this whole "AI" thing is overblown?
🤦♂️
Nvidia: We're guiding to 70% revenue growth next year. There's incredible AI demand for 100% growth, but we can't supply it.
Google: Everyone wishes they had DOUBLE the compute capacity right now. There's so much AI demand.
Bloomberg/FT: There's a $500 million market-cap Korean neocloud no one has ever heard of and it traded poorly on its IPO first day. AI IS DEAD. EVERYONE FREAK OUT AND PANIC!