Summary: A new Geekbench 6 leak attributes to the Exynos 2600 — Samsung’s first 2 nm GAA SoC — a score of 4,217 in single-core and 13,482 in multi-core. The striking detail: the performance per thread matches Apple M5 (≈ 4,263), something unprecedented in Android phones. The sample is engineering, and as a precaution, the result is no longer present in the public Geekbench database (it may have been removed or manipulated), so it’s advisable to treat it as provisional.
What does the leak say
Deca-core CPU (1+3+6)
1 “prime” at 4.20 GHz
3 performance cores at 3.56 GHz
6 efficiency cores at 2.76 GHz
Geekbench 6 scores
Single-core: 4,217 (≈ +22% compared to a previous pass of 3,455)
Multi-core: 13,482 (≈ +16% versus 11,621)
Declared comparison
Apple M5: single4,263 (≈ +1% vs. this Exynos), multi17,862 (≈ +32% vs. this Exynos)
The source also recalls a efficiency data point from previous leaks: 7.6 W of power in multi-core (Geekbench 6) and -59% power consumption compared to an A19 Pro. There’s no public evidence of these figures in the current leak, so treat with caution.
Exynos 2600 Leak: Engineering Sample Matches Apple M5 in Single-Core (Geekbench 6) and Reaches 4.20 GHz in Its Top Core 4
Why it could be significant (if confirmed)
Historic jump in single-core: Android has lagged behind Apple’s core for years. If an Exynos matches the M5 in single thread, UX performance (latency, app opening, system response) would improve significantly.
Pico frequency of 4.20 GHz: an unprecedented mark for Samsung on mobile. Achieved with 2 nm GAA, promising better leak control, but thermal sustainability in a slim chassis remains to be seen.
Multi-core plans: the M5 would stay ahead in sustained load (multi +32%). It appears the 2600’s priority is improving the critical thread without sacrificing multi-core gains.
Open questions (and why caution is advised)
Verification: the run is not currently in the public Geekbench 6 database. It may have been removed or could be fake.
Profiles and governors: an engineering sample could use aggressive voltage/frequency settings that won’t appear in retail software.
Actual efficiency: without comparable power curves (TTF, p-states, thermal limits), it cannot be confirmed that the final mobile will perform the same or better at lower W.
GPU/NPU: the leak focuses on CPU. For on-device AI (translation, image expansion, local assistants), the NPU and memory subsystem (LPDDR6?) will be critical.
What might be under the hood
2 nm GAA process (Samsung Foundry): the shift from FinFET to Gate-All-Around aims for better leak control and lower Vmin. If binned properly, 4.20 GHz in burst could be plausible.
Microarchitecture: no official details, but the 1+3+6 strategy resembles a very large prime core (high IPC) + three performance cores for medium loads + six efficiency cores for background tasks.
Memory and cache: much of modern single-core performance depends on cache hierarchies and fine-tuned prefetchers; not just raw frequency.
Calendar and context
Galaxy S26: expected in February 2026. Exynos 2600 would be the candidate for Exynos markets (Europe/Korea), with Snapdragon in others, as usual.
Competitors: A19 Pro/M5 from Apple, Snapdragon X/8 Gen from Qualcomm, and advancements from MediaTek (Dimensity) will shape a competitive 2026.
AI landscape: the market is leaning toward longer cycles and local models; CPU matters for orchestration, but NPU/GPU will drive AI demos on device.
What it means for users (if it arrives as is)
Seamless experience: animations, scrolling, and “instantaneous” loads closer to Apple’s feel.
Gaming and heavy apps: faster compile times and lower frame time spikes; thermal sustainability remains to be confirmed.
On-device AI: will depend on the NPU. Expect compatibility with MCP/NIM-like runtimes from the Android ecosystem and acceleration of transformers with KV-cache.
Conclusion
The leak of the Exynos 2600 indicates that Samsung may finally match Apple in single-core for the first time in years, thanks to a “prime” core reaching nearly 4.20 GHz on 2 nm GAA. Nevertheless, public verifiable evidence, power figures, and especially thermal tests on final hardware are still missing. If the scores are confirmed in retail devices, 2026 could be the year Android closes the single-thread gap and elevates on-device AI… but caution remains until then: it’s an engineering sample, and the database does not currently support it.
