A Kizilelma CCA conducts the world's first autonomous air-to-air engagement with a live missile.

Kizilelma: Turkey hits major air combat milestones

The Kizilelma collaborative combat aircraft from Turkey’s Baykar Technologies has hit two major air combat milestones as the competition to field a functioning autonomous fighter heats up. 

BLUF: 

  • In November, the Kizilelma conducted a live air-to-air missile engagement, believed to be a world first. 
  • In December, two of the aircraft flew in formation, which Baykar states is a world first. However, Lockheed’s Skunk Works did conduct a flight with two uncrewed aircraft controlled by a piloted aircraft in November 2024. 
  • The demonstrations are an encouraging sign that the collaborative combat air capability sought by multiple nations is achievable. 

In the world of collaborative combat air (CCA) platforms, there are many technological hurdles to overcome. Amongst them are the general challenges of building an airworthy, jet-powered aircraft that can fly at a similar speed to a fighter jet. But there are many others relating to the required autonomy that will decide whether or not CCAs live or die as a concept. And that is why the late-2025 tests by Baykar Technologies with the Kizilelma (pronounced KUH-zuhl-EL-mah) are so interesting. In two separate tests, one reported on November 29, and the second on December 27, the Kizilelma engaged a target aircraft with a beyond visual range air-to-air missile and flew in formation autonomously. The Turkish CCA is not alone in some of these feats, indicating that the world is coming closer to realising autonomous air combat. 

In the air-to-air demonstration, the target jet was illuminated by Kizilelma’s Murad AESA radar from Aselsan, and engaged with a Gökdoğan missile from Tubitak Sage. Tubitak Sage is Turkey’s national research and development centre that cooperates with the country’s Ministry of Defence and defence enterprises on important projects including precision strike missiles, guidance systems, and of course air-to-air missiles. The Gökdoğan has a range exceeding 65 km with a radar seeker with a datalink that allows for mid-course updates and something called home on jamming, which is where the missile targets any jamming emitted by the target rather than its radar returns. It’s a neat capability that gives the missiles two channels to find their target. 

During the test, the Kizilelma flew with a formation of five F-16s, which carried additional air-to-air missiles and the commander of Turkey’s air force, Gen. Ziya Cemal Kadıoğlu, as well as the general manager of Aselsan and chairman of Baykar Technologies. Flying in formation with crewed aircraft is another important area of development for CCAs, and one that has been explored by Lockheed Martin and Boeing, amongst others. 

Lockheed’s Skunk Works flew a mixed offensive counter air mission in 2024, with a crewed aircraft controlling two uncrewed, AI-piloted aircraft to engage a simulated target. Throughout a series of tests, the battle commander aboard an L-39 Albatross issued real-time commands through a touchscreen pilot vehicle interface, assigning targets to two AI-piloted L-29 Delfin jets which then worked collaboratively to defeat simulated adversary aircraft.  

Then, in June this year, Boeing and the Royal Australian Air Force (RAAF) demonstrated the interoperability of the MQ028 Ghost Bat with an E-7A Wedgetail airborne early warning and control aircraft. The demonstration, part of the Capability Demonstration 2025 series, saw a single operator aboard an airborne E-7A Wedgetail effectively control two in-flight MQ-28 Ghost Bat aircraft, alongside a third digital MQ-28. This capability validates a core tenet of the Ghost Bat’s operational concept: extending the reach and protective umbrella of crewed assets by acting as forward scouts or escorts.

Two prototype Kizilelma aircraft fly in close formation using an autonomous flight pilot.

Two prototype Kizilelma aircraft flew in close formation using an autonomous flight pilot. Credit: Baykar Technologies

Baykar did not get to enjoy its unique status for long, however. The Ghost Bat soon followed, with a 12 December press release from the RAAF stating that an MQ-28A had become “one of the first collaborative combat aircraft (CCA) to successfully engage an aerial target with a live weapon.” During that test, the Australian CCA was acting as a loyal wingman to an E-7A, and engaged the target, a Phoenix uncrewed aerial target, with an AIM-120 missile. The team working on that project went from idea to test in nine months, the press release states. 

General Atomics has simulated air-to-air weapons release with its MQ-20 Avenger, which forms the technology development model for its other work. During those tests, in June 2025, the aircraft flew a patrol mission and collaborated with crewed and uncrewed aircraft to simulate the engagement of two aircraft.

So it appears that there are elements of Baykar’s second announcement that have been replicated in other trials. Announced on December 27, 2025, the two prototype Kizilelma aircraft flew in close formation using an autonomous flight pilot. This was also a world first, according to the press release, but this seems harder to assess for certain. There do appear to have been other formation flights, but it is likely the degree of autonomy that was unique. 

As part of that test, the aircraft took off autonomously and flew a combat air patrol without commands from a pilot. “In the future, patrol and interception missions—traditionally conducted by manned fighter aircraft to secure designated airspace—will be performed autonomously by Kizilelma fleets,” the press release states. While Baykar claims a world first for fully autonomous close formation with jet-powered UCAVs, it is worth noting that the DARPA ACE program and Lockheed’s 2024 tests achieved similar milestones with ‘human-in-the-loop’ or simulated parameters.

All in all, it is clear that the closing days of 2025 brought with them some interesting developments in the world of CCAs. These tests indicate that the technologies involved are maturing and that the political will to make them a success is strong. However, it is also interesting to note that Australia and Turkey appear to be ahead of the US and European states, which are essentially at much earlier stages of CCA development. 

Calibre comment: How do AI pilots perform against humans in combat?

AI pilots have been tested in a range of scenarios including simulations and real world flights with some interesting results. One of the most well-known examples is the AlphaDogfight trials conducted by Heron System (now part of Shield AI) in 2020. In those simulated tests, an AI pilot beat the pilots of seven other companies and flew against an experienced US Air Force pilot, winning 5-0. The goal was to demonstrate that AI pilots could handle tasks like dogfighting, allowing human pilots to manage the combat, rather than conduct it. The other notable trial is the DARPA 2024 trial where an AI pilot flew a real F-16 against a piloted F-16 in a within visual range scenario. The AI pilot flew very precise manoeuvres so the safety pilots on the autonomous F-16 felt no need to take control, and the score was reportedly even. 

What all of these limited tests show is that using AI pilots is not just possible, but likely to be tactically desirable. Pilots are a major bottleneck for most air forces. They are expensive to train, and maintaining their skills at a high level requires regular flight hours. Now consider that in 2016, the US Air Force said that flying 200 hours a year enabled pilots to get a lot better at everything, and reported an average of 150 flight hours per pilot per year. An F-35 is regularly said to cost $30,000 or more per hour of flight, which means just meeting the flight costs of an F-35 pilot for one year (not including salaries etc) would come to $4.5 million. Per pilot. CCAs do not need such a regular drum beat of flight training to meet their capabilities and do not need pilots, so the costs can be significantly lower whilst theoretically allowing for force expansion, if the one to many paradigm can be solved. 

By Sam Cranny-Evans, published on January 8, 2026. The lead image shows the Kizilelma firing its air-to-air missile and the moment that the missile hit its target. Credit: Baylar Technologies. 

Get insider news, tips, and updates. No spam, just the good stuff!