Fordel Studios
We are in the densest period of AI model releases in history. Every week resets the leaderboard. This is the scorecard as of March 30, 2026 — what matters, what moved, and who is bluffing.
Who Is Actually Leading the AI Race Right Now?
Scorecard: ranked on reasoning quality, coding performance, multimodal depth, agentic capability, and enterprise traction. Trend reflects last 30 days.
- 1. Anthropic (Claude 4.x) — 9 / 10 | Trend: UP — Best-in-class coding agent. Claude Code is now the default choice for agentic dev workflows. Sonnet 4.6 punches above its weight on price.
- 2. OpenAI (GPT-5 / o3) — 8.5 / 10 | Trend: FLAT — Still dominant on ecosystem size and enterprise sales. o3 leads on hard reasoning. Codex got a plugin system this week for enterprise governance. But the gap with Anthropic on coding has narrowed visibly.
- 3. Google DeepMind (Gemini 3.1) — 8 / 10 | Trend: UP FAST — Gemini 3.1 Flash Live shipped this week with real-time multimodal streaming. Google is the most dangerous player nobody is talking about enough. The Live API changes what agents can do with voice and video.
- 4. Meta (Llama 4 series) — 7.5 / 10 | Trend: UP — Open weights, runs on your infra, no per-token bill. Llama 4 Scout is genuinely competitive on mid-tier tasks. Threatening the entire cloud-hosted mid-market.
- 5. xAI (Grok 3) — 6.5 / 10 | Trend: UP — Real-time data from X/Twitter is a durable moat. Grok 3 is the only frontier model that knows what happened this morning. Enterprise sales still nascent.
- 6. Mistral (Large 2 / Codestral) — 6 / 10 | Trend: FLAT — The European option. Codestral is solid for constrained coding tasks. Not winning on raw benchmark numbers but winning on GDPR compliance and data residency conversations.
What Were the Top 3 Moves This Week?
Three things that actually shifted the score this week:
Move 1 — Google ships Gemini 3.1 Flash Live
Real-time conversational agents with live video and audio. This is not a benchmark story — it is a product story. The Live API lets you stream sensor data, camera feeds, and voice input to a model mid-session. That changes what enterprise automation looks like. Google buried the lead on this one.
Move 2 — OpenAI adds a plugin system to Codex
Enterprise governance for AI coding agents: audit trails, policy controls, integration hooks. On paper this sounds boring. In practice it is OpenAI answering the one question that blocks enterprise procurement: how do we control what the agent does? Smart move. Directly targets JPMorgan-tier buyers who were told last week to use AI or fall behind.
Move 3 — Two-thirds of all code is now AI-generated
TechRadar published this figure this week. That is not a prediction — that is a current measurement. Every model company now competes on who owns the AI-generated code layer. Anthropic is winning that fight today with Claude Code. OpenAI is fighting back with Codex plugins. Google is not in this conversation yet, and that is a visible gap.
Is Gemini Closing the Gap Faster Than Anyone Admits?
Yes. And the people most surprised will be the ones anchored on 2024 benchmarks.
Gemini 2.0 and 2.5 were good. Gemini 3.1 Flash Live is different in kind, not degree. Google has distribution advantages that no one else has: Android at 3B devices, Google Workspace at 3B users, YouTube, Search, Maps. When Google ships a capable model it does not need to win a benchmark — it just has to be good enough and be everywhere.
The coding agent story is still Anthropics. The enterprise reasoning story is still a three-way fight. But the multimodal, real-time, embedded-in-everything story? Google is winning that right now and it is not close.
“Gemini is closing the gap faster than anyone admits. Google does not need to beat Claude on coding. It needs to be the default model for everything else.”
What Does the Bottom Line Look Like?
- Anthropic: The coding agent company that also makes frontier models. Claude Code is their moat. Do not underestimate Sonnet 4.6 on cost-performance.
- OpenAI: Still the default enterprise shortlist entry. Ecosystem and brand do enormous work. But they are reacting more than leading right now.
- Google: The most underrated player in the enterprise AI race. Flash Live is a category move. Watch the next 60 days.
- Meta: The open-source threat nobody is pricing into their vendor contracts. If your use case fits, Llama 4 just made your per-token bill optional.
- xAI: Real-time data is the only moat here. Narrow but real. If your product needs to know what happened today, Grok is the only call.
- Mistral: Not winning on performance. Winning on compliance conversations in regulated European markets. That is a real niche.
The race is not over. It is not even at halftime. Every quarter from now produces a new scorecard. What I can say is this: the teams shipping product capabilities — not benchmark press releases — are the ones moving the score. Right now that is Anthropic and Google. OpenAI is managing a lead. Everyone else is playing for second.
Check back in 30 days. This scorecard will look different.
