😺 Gemini 3.8 and Muse Spark 1.3 go head to head for third place
The Neuron Eric Gerard Ruiz ● Covered by 2 sources
Google and Meta both launched new workhorse AI models on the same day. The fight is really about which one gets a real job done cheaper and faster.
Based on reporting by The Neuron, Eric Gerard Ruiz — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google and Meta picked the same day to push out new AI models, and both were aimed at the boring-but-important middle of the market: the fast, cheap systems companies can actually run all day. Google released Gemini 3.8 Flash. Meta released Muse Spark 1.3.
These weren’t the biggest models either company has. They were built for coding, research, and AI agents, which means the kind of tasks that don’t stop at one answer. Gemini 3.8 Flash was updated to handle coding, reasoning, and long-running work better. Muse Spark 1.3 was tuned to follow instructions more closely, use tools more carefully, ask questions, and know when to stop.
Meta says Muse used about 20% fewer tool calls and 25% fewer tokens than its predecessor. Independent tests put the two models close, with Muse ahead on some intelligence tests and Gemini much faster. So this isn’t a simple race to the top. It’s a contest over how much work an AI model does before it decides it’s done.
That matters because agentic systems burn money step by step. Give one an assignment like researching companies and building a spreadsheet, and it may search, read, run code, check itself, fix mistakes, and loop again. Google seems to be betting on heavier thinking when things get hard. Meta seems to be betting on smarter restraint.
There’s one more twist: Google kept Gemini’s token price the same as the previous version, but Artificial Analysis found it cost about 40% more per completed task because it did more work. That’s the real metric now. Not just what a model costs per token, but what it costs to finish the job correctly.
My take — AI-written commentary, not fact-checked reporting
This is the right fight. The industry has spent too long bragging about raw intelligence like it’s a leaderboard at a science fair. The unglamorous truth is that the model that finishes cleanly, with fewer wasted steps, is the one companies will keep paying for.
Read more about this at: The Neuron