What changed
IBM Quantum opened early access to ibmberlin for Premium and Flex plans. The machine is the first Nighthawk r1 QPU in the European eu de region. It follows ibmmiami in the United States. Public docs describe 120 qubits on a grid style layout.
IBM states median two qubit gate times near 68 ns on ibmberlin. That compares with about 138 ns on ibmmiami. Coherence targets stay in a similar band, with T1 near 350 µs in the same update. Qiskit added a FakeBerlin simulator backend in April 2026 so teams can test circuits before burning queue time.
IBM still labels ibm_berlin as exploratory. The default repetition time is 4 ms instead of the usual 250 µs unless users pick a shorter value. Dynamic circuits are not supported yet. Mid circuit measurement quality is limited while the system is tuned.
Why it matters
EU based labs often need data residency and lower network latency to a QPU. A Berlin region Nighthawk gives them a local option without routing every job to the US east fleet. Faster two qubit gates can shrink depth for lattice style workloads that Nighthawk targets.
The tradeoff is real queue cost. A longer default repetition time can inflate billed QPU minutes for the same circuit depth. Teams must read limits before they move production chemistry or optimization studies.
Who is affected
Quantum algorithm leads on IBM Premium or Flex contracts in Europe. Cloud architects who map workloads to regions. Partners who sell hybrid classical quantum workflows to automotive and materials clients. Students on lower tiers may still wait for wider access.
What to do next
Read the IBM Quantum latest updates note on ibmberlin. Compare gate times and repetition defaults against your current ibmmiami baselines. Run transpiled benchmarks on FakeBerlin in Qiskit IBM Runtime before you request large session blocks.
What to watch
IBM lifting the 4 ms default repetition cap without a quality warning. Support for dynamic circuits on ibm_berlin. Public papers that cite EU Nighthawk runs at scale, similar to recent US fleet work.
Image: GitHub
How Nighthawk fits the IBM fleet
Nighthawk builds on Heron era couplers and signal delivery but adds grid connectivity. That helps algorithms that map naturally to two dimensional qubit grids. IBM positions the line for regimes where classical simulation gets costly.
Early access is not the same as a stable production SLA. Exploratory flags mean calibrations and limits can change week to week. Treat first results as directional, not as a final benchmark for grant reporting.
Partners should document which backend version and repetition setting each run used. That habit saves pain when IBM updates defaults during tuning cycles.
Limits teams should plan for now
No dynamic circuits means you cannot branch on mid run measurement outcomes on ibm_berlin today. If your roadmap needs feed forward control, stay on backends that already support it or stay in simulators until IBM clears the flag.
The 4 ms default repetition time can dominate wall clock cost for shallow circuits. IBM lets you pick a shorter repetition window, but warns that error rates may rise. Run small A/B tests before you lock a campaign budget.
Grid topology is not fully connected. You still need smart routing and swap overhead in transpilation. Use FakeBerlin to stress test layout before you claim a speedup from gate time alone.
If your legal team cares about region, note that eu de hosting is the main draw. The physics win is real, but the ops win is local access with clear data handling rules.
Image: GitHub
Sources
- Primary — IBM Quantum (via doQumentation mirror), Latest updates including Nighthawk in the EU (2026). Documents ibm_berlin early access, gate times, and temporary limits.
- Primary — Qiskit, Add fake Berlin backend commit (April 2026). Confirms 120 qubit FakeBerlin configuration tied to ibm_berlin.
Related in Quantum
Continue reading