Inkling 2 Mini vs Beam 1.0 501B-A23B NVFP4

Change models
Inkling 2 MiniThinking Machines Lab
IndexAll categories, out of 100.
57.7 ±4.2 best
57.6 ±5.4
GapIndex points below the top.
Top
−0.1 · no clear gap
CoverageIndex weight that has results.
90% best
85%
SpeedOutput tokens per second.
88/s
—
EloLMArena rating and rank.
—
—
ReleasedMonth of release.
2026-11
2026-10
Input / 1MUS dollars per 1M input tokens.
$0.10
—
Output / 1MUS dollars per 1M output tokens.
$0.40
—
1K chats2K tokens in, 500 out, each.
$0.40
—
1K documents20K tokens in, 1K out, each.
$2.40
—
1K agent tasks200K tokens in, 20K out, each.
$28.00
—
ContextMaximum tokens in one request.
1.05M best
1M
AgenticMulti-step tasks with tools.
60.5 best
54.3
CodingCode writing and repair.
60.0
60.0
ReasoningLogic problems and puzzles.
61.6
61.6
MultimodalTasks with images and text.
57.1
—
KnowledgeFacts and expert knowledge.
49.6
58.5 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
—
61.9
MathMath problems.
47.4
54.8 best

Benchmarks

9 shared
Only Inkling 2 Mini: 27Only Beam 1.0 501B-A23B NVFP4: 9
Inkling 2 Mini
Beam 1.0 501B-A23B NVFP4
79.2%24 / 114
80.7% best15 / 114
71.6%51 / 89
78.4% best41 / 89
86.6%53 / 87
89.3% best38 / 87
DeepSWEAgentic
48.2% best56 / 64
42.7%61 / 64
BrowseCompAgentic
79.0% best31 / 63
75.6%39 / 63
30.7%41 / 55
34.8% best34 / 55
MCP AtlasAgentic
82.7% best12 / 53
80.5%16 / 53
85.3% best12 / 25
81.1%14 / 25
LongBench v2Reasoning
66.9%5 / 25
67.2% best4 / 25