Personalized cancer vaccines have been the “almost” story of immuno-oncology for about a decade now.

The premise is appealing. Sequence the tumor. Find mutations that make it look different from normal tissue. Design a vaccine that trains the patient’s own immune system to attack those specific mutations. Give it alongside a checkpoint inhibitor. Watch the combination do what neither one can do alone.

Moderna and BioNTech are both deep into Phase 2/3 programs for this. BioNTech’s autogene cevumeran recently posted Phase 2 data in adjuvant melanoma. Moderna’s mRNA-4157 with Merck’s pembrolizumab is moving toward Phase 3 in multiple tumor types.

The smaller players tend to get overlooked in that conversation.

Evaxion has been building quietly in Copenhagen for a few years. Yesterday at AACR 2026, they dropped Phase 2 data that deserves more attention than it’s probably going to get.

EVX-01 is a personalized peptide-based neoantigen vaccine, designed by Evaxion’s AI-Immunology platform. It’s given in combination with Merck’s Keytruda (pembrolizumab) in first-line advanced melanoma.

The headline earlier results showed a 75% objective response rate. Yesterday, they reported updated Phase 2 data focused on two things that matter more than the ORR number alone.

First: 86% target precision. That means 86% of the peptides the AI picked as neoantigens produced tumor-specific immune responses when injected into patients. For AI-designed therapeutics, that’s the real metric. Not how pretty your model output looks. What percentage of your predictions survive contact with human biology.

Second: immune responses sustained over two years. Durability has been a question for neoantigen vaccines. You can train the immune system to recognize something. Whether it remembers the lesson over years is a different question. The Evaxion data suggests the lesson sticks.

Let’s be honest about what Evaxion is. They’re publicly traded on Nasdaq as EVAX. Small market cap. The company has had to deliver on the AI-Immunology platform because the alternative, being another small AI-drug-discovery company with a platform story and no data, has not been kind to anyone over the last 18 months.

What yesterday’s data does is change the conversation. It moves Evaxion from “interesting AI vaccine platform” to “AI platform that predicted immunogenic targets at 86% precision and whose responses lasted two years.”

Those are the two weak points that every AI drug discovery pitch has to defend. Evaxion defended both.

Additional long-term clinical readouts from the trial are expected later in 2026. If they hold up, this positions EVX-01 as a genuine contender in the personalized cancer vaccine space, a space that’s attracting massive AI investment right now but still needs proof points that the predictions work in human patients.

Evaxion has one.