Building @IzwiAI - Private Voice AI Runtime
Harare, Zimbabwe
Joined July 2009
- Tweets18.3K
- Following1.6K
- Followers1.9K
- Likes19.9K
Pinned Tweet
300+ GitHub stars for @IzwiAI
1200+ Downloads π
I strongly believe voice AI will be huge in the next few years.
Izwi's goal is to be the default voice AI runtime for private deployments.
github.com/izwi-ai/izwi
Lennex Zinyando retweeted
π Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. Weβre sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
π github.com/deepseek-ai/DeepIβ¦
1/4
Lennex Zinyando retweeted
I wanted one Mac window on my iPad, not my whole desktop, so I built Porthole.
Pick a window and it streams over your WiβFi at up to 120fps. Use it like an iPad app: tap to click, drag to scroll, pinch to zoom. Works on iPhone and Android too.
github.com/iamngoni/porthole
Lennex Zinyando retweeted
π DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous versionβa Pareto improvement in performance and cost. Now live on our website, app, and API.
π οΈ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
Lennex Zinyando retweeted
Today weβre releasing Muse Spark Open, the most capable US open-weight model to date.
Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.
Key Model Specs:
β 816B total / 42B active parameters
β 1M token context window
β Natively multimodal across text, image, video, audio
Lennex Zinyando retweeted
Mistral is back:
-pushing the frontier on dollar per intelligence (1.13$ per task for a 38. mimo v2.6 pro gets 46 for 0.13$)
-best open model out of us or europe (competition: nemotron 3 ultra 23, gemma 4 17, gpt-oss 12)
-same score as gpt 6 luna. luna costs 0.07$ per task. 16x
-#64 of 225 on artificial analysis. frontier
-1 trillion parameters to land 3 points under GLM 5.3 flash (320B, 0.25$)
-open weights (AA lists it as proprietary. weights oct 27)
-most open license yet (large 3 apache 2.0, medium 3.5 modified mit, this one "custom mistral license")
-only 8 points behind the best chinese open model and 20 behind opus 5.5
-beat glm 5.3 on deepswe, 62 to 61 (on mistrals chart. the live leaderboard has glm 5.3 at ~69)
-state of the art in legal (15%)
-highest cyber score of any model, 82% (opus 5.5 and astra refuse the task and score near zero)
-same blog post brags it refuses more cyber prompts than any other open model
-200M output tokens to run the index, median is 81M. le chonk talks
-+29 points over large 3 (large 3 had 9)
-sovereign ai (they resell glm 5.3 since august. until today the smartest model on mistrals own platform was chinese)
-3 billion euro raised, 21 billion valuation, first model out of it: 38
-"no signs of saturation" (the rl run isnt finished, they shipped anyway)
-the meme promised 30 trillion parameters. we got 1 (49B active)
-europe can compete, said mensch. in abu dhabi
-2nd of 5 in their own blind eval. behind opus 5. not 5.5
Lennex Zinyando retweeted
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.
GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.
Weβve also made caching and inference more efficient, and weβre passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPTβ5.6 promotional pricing.
Lennex Zinyando retweeted
π¨ Introducing π π’π₯ππ‘π
Today, we are releasing π π’π₯ππ‘π, a new multilingual AI model built for African languages.
π π’π₯ππ‘π is a 1.5 billion parameter language model trained from scratch for 12 African languages, with its own tokenizer designed for Africa.
$40,000 worth of GPUs
π§π΅πΆπ πΆπ π π’π₯ππ‘π: the largest African language model ever trained from scratch, at 1.5 billion parameters. Not a fine tune, this was trained from the ground up for Africa. π
Trained on 12 African languages, it beats models from Google, Meta, Alibaba and HuggingFace on African language modelling and translation while being up to 8x smaller.
π π’π₯ππ‘π doesn't belong to us. The weights are open and free to use commercially, so researchers can build on a base that doesn't treat African languages as an afterthought.
Weights in the comments.
Thanks to @UNDP, @AIHub4SD and CINECA for the GPUs and the support. None of this happens without them.
You think you know pain? Try getting GPU quota increases on @googlecloud ππ
Lennex Zinyando retweeted
I've been using Astra all day, and it's definitely a step forward compared to Sol. It makes fewer mistakes, understands my codebase better, and asks more clarifying questions. Now waiting for @thsottiaux to give us another reset for the weekend so I can continue playing π
One of the goals of @IzwiAI is to democratise LLM inference. Organisations should be able to host and serve LLMs efficiently and cheaply to serve their needs. We are starting with voice and smaller models that are not supported widely. More in the next few weeks
Thatβs what we are doing @IzwiAI, building voice models that understand African languages.
Finally, a banked reset ππΎππΎ
It's me again. I come bearing great news.
First of all, we have hit 20M active users for Codex some time this week. Second of all, this is cause for celebration and during the day we will credit every Codex and ChatGPT Work user with a BANKED reset that you can use at your own leisure. And we will have some other good news later too!
Now, on usage limits draining faster, while we're not seeing anything abnormal, we do take it incredibly seriously and there is an ongoing investigation. I will share if we do find anything and my below post is really a clarification on a specific pattern that we did see that I wanted to call out.
Go do something amazing today.
Local first voice AI for speech, chat, and audio workflows.
github.com/izwi-ai/izwi
