Beam 1.0 vs DeepSeek-V3.2
Wins 6 of 6 areas
Coding · Agents · Reasoning · Math · Long documents · Following instructions
Wins 0 of 6 areas
—
Beam 1.0 is the stronger all-rounder.
Scores updated · 17 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- CodingWriting and fixing software50Beam 1.05 of 5 tests
- ReasoningHard problems that need careful thinking40Beam 1.04 of 4 tests
- AgentsCarrying out multi-step tasks on its own20Beam 1.02 of 2 tests
- MathCompetition and research-level math10Beam 1.01 of 1 test
- Long documentsFinding answers in very long texts10Beam 1.01 of 1 test
- Following instructionsDoing exactly what it is asked10Beam 1.01 of 1 test
Math, long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Beam 1.0 pulls ahead
- Long, multi-file coding tasks in real codebasesSWE-bench Pro+50.5points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+33.3points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+25.4points ahead
Where DeepSeek-V3.2 pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 17 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingBeam 1.0
- SWE-bench ProBeam 1.0 by 50.566.115.6+50.5
- Terminal-Bench 2.1Beam 1.0 by 33.380.146.8+33.3
- SciCodeBeam 1.0 by 10.349.238.9+10.3
- SWE-bench VerifiedBeam 1.0 by 8.581.673.1+8.5
- SWE-bench MultilingualBeam 1.0 by 8.178.370.2+8.1
AgentsBeam 1.0
- BrowseCompBeam 1.0 by 25.476.851.4+25.4
- MCP AtlasBeam 1.0 by 18.680.862.2+18.6
ReasoningBeam 1.0
- ARC-AGI-2Beam 1.0 by 19.523.54+19.5
- CritPtBeam 1.0 by 13.416.32.9+13.4
- Humanity's Last ExamBeam 1.0 by 11.43624.6+11.4
- GPQA DiamondBeam 1.0 by 6.690.684+6.6
Following instructionsBeam 1.0
- IFBenchBeam 1.0 by 18.979.660.7+18.9
Other results3 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- DeepSearchQA (F1)Beam 1.0 by 19.680.560.9+19.6
- ARC-AGI-1Beam 1.0 by 177457+17
- HLE (with tools)Beam 1.0 by 10.351.140.8+10.3
Questions people ask
Which is better, Beam 1.0 or DeepSeek-V3.2?
Beam 1.0 wins all six areas where both have results: coding, agents, reasoning, math, long documents and following instructions. DeepSeek-V3.2 wins none.
Which is better for coding?
Beam 1.0. It wins 5 of the 5 coding tests both models report; DeepSeek-V3.2 wins none.
How do you compare the two?
We use the 17 benchmark tests both models have published scores on. The verdict counts the 14 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 3 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.