Mrinal Garg

Building @via_ai_ | prev IITB

[Via @via_ai_]

Today we’re coming out of stealth to give enterprises frontier AI coding at half the cost.

Via is a middleware that sits between a CLI / IDE and an LLM, installs in 10 seconds, and has already demonstrated a new standard of performance.

- On DeepSWE 1.1, Claude Code (Opus 5.5) is 306% the cost of Via US (US-model-only configuration). Via US is equally accurate.

- On Terminal Bench 4.0, Claude Code (Opus 5.5) is 150% the cost of Via. Via is more accurate.

Via runs inside your VPC, so no data leaves your environment.

We’re in production at a public healthcare company, with trials underway at several of the largest tech companies in Silicon Valley. We have room for five more enterprises before the end of the year: [email protected]

Our co-founder @ilaamathur on why we built Via:

[Mrinal Garg @MrinalGarg15569]

@via_ai_ Let’s gooo!!

[Mrinal Garg @MrinalGarg15569]

I am so grateful to be a part of this team. The intuition, research and love for building is phenomenal on this team. Working alongside @ilaamathur , @ShrayAlag and @rishi_deshpande and the rest of the team is such a blessing!

[Shray Alag @ShrayAlag]

@via_ai_ is the first product to provide frontier AI coding accuracy at half the cost.

Claude Code is 306% the cost of Via at the same accuracy on DeepSWE v1.1. Check out Ila’s post below for more detailed benchmarks.

We’re already in trials at several of the largest tech companies in Silicon Valley and are in production at a public healthcare company.

Via is powered by Codegraph (our LLM-optimized compression algorithm that led us to be #1 on SWE-Bench Verified for several months last year) and our new Umami agent architecture.

Via integrates with your favorite IDE/CLI in one terminal command, is deployed within your VPC, and has an unchanged dev experience.

We are a small team of Olympiad camp qualifiers (with multiple gold medals), researchers, and repeat founders, backed by leaders like Jeff Wilke (ex-CEO Amazon Worldwide Consumer), Manik Gupta (ex-Uber CPO), and others from Anthropic, Google, Uber, Apple, Snowflake.

Via is designed for enterprises with 7 figure coding budgets and above. If this applies to you, DM me or email [email protected] and we’ll get back to you very soon.

Stay tuned for a deep dive into Codegraph and Umami.

[Mrinal Garg @MrinalGarg15569]

@ilaamathur @via_ai_ @jeffawilke @ShrayAlag Excited to be part of the journey till now and thrilled for what’s coming next!

[Mrinal Garg @MrinalGarg15569]

I am excited to announce that @via_ai_ is out of stealth!

I couldn’t be prouder of the research and hard work this team has done to be at the frontier of AI coding accuracy while dropping the cost to half.

More from our cofounder @ilaamathur

[Ila Mathur @ilaamathur]

Today, @via_ai_ is coming out of stealth to provide enterprises frontier AI coding at half the cost.

We are a team of Olympiad camp qualifiers (with multiple gold medals), researchers, and repeat founders from Stanford and IIT.

Via is backed by @jeffawilke (ex-CEO Amazon Worldwide Consumer), @manikgupta (ex-Uber CPO), @shaileshrao (ex-Google MD), and other legends from Anthropic, Apple, Snowflake, and Twitter.

@ShrayAlag and I left Stanford CS to build Codegraph, an LLM-optimized compression algorithm. Codegraph put Via at the top of SWE-Bench Verified for several months of 2025, ahead of Anthropic, OpenAI, Cognition, and Cursor (single model, no test-time compute).

Now, we’re bringing this research, alongside our new Umami agent architecture, to market. Via integrates in 10 seconds with any CLI / IDE and runs within your enterprise’s VPC.

On benchmarks, Via and Via US (US-only models) set a new standard of performance:

- On DeepSWE v1.1, Claude Code (Opus 5.5) is 306% the cost of Via US. Via US is equally accurate.
• Devin Fusion is 244% the cost of Via US, Cursor Router is 294%.

- On Terminal Bench 4.0, Claude Code (Opus 5.5) is 150% the cost of Via. Via is more accurate.
• Devin Fusion is 155% the cost of Via, Cursor Router is 134%.

We are in production at a public healthcare company, with trials underway at several of the largest tech companies in Silicon Valley.

We have room for 5 more enterprises before the end of the year. DM me or email [email protected] and we’ll get back to you very soon.

See the comments for how Via would deploy for your team.

[Ila Mathur @ilaamathur]

McKinsey reports that 93% of enterprises have exceeded their AI budgets. Spend is only increasing

Getting more out of AI shouldn’t mean paying more for it. Open source models and Fable/Astra alternatives offer much more intelligence per dollar. But routing between them can break prompt caching and disrupt the developer experience.

Our team of Olympiad medalists, ex-founders, and researchers has spent the last year building algorithms to cut coding costs without sacrificing frontier accuracy or the experience developers love.

More soon

https://t.co/1FkZ6V3Wbq

[Mrinal Garg @MrinalGarg15569]

@heynavtoor The paper is from September 2025. We gotta stop these posts that put “BREAKING!!” on literally anything