Inference
EN

Claude Haiku 5.5 vs Inkling 2 Small

Change models
Claude Haiku 5.5Anthropic
Inkling 2 SmallThinking Machines Lab
IndexAll categories, out of 100.
66.3 ±2.8
66.4 ±3.8 best
GapIndex points below the top.
−0.1 · no clear gap
Top
CoverageIndex weight that has results.
95% best
90%
SpeedOutput tokens per second.
159/s
—
EloLMArena rating and rank.
1399
1403 best
ReleasedMonth of release.
2026-10
2026-11
Input / 1MUS dollars per 1M input tokens.
$0.10 best
$0.30
Output / 1MUS dollars per 1M output tokens.
$0.50 best
$1.00
Batch output / 1MBatch price per 1M output tokens.
$0.25
—
1K chats2K tokens in, 500 out, each.
$0.45 best
$1.10
1K documents20K tokens in, 1K out, each.
$2.50 best
$7.00
1K agent tasks200K tokens in, 20K out, each.
$30.00 best
$80.00
ContextMaximum tokens in one request.
1M
1M
AgenticMulti-step tasks with tools.
68.6 best
67.6
CodingCode writing and repair.
71.0 best
66.6
ReasoningLogic problems and puzzles.
67.8
68.4 best
MultimodalTasks with images and text.
51.9
63.6 best
KnowledgeFacts and expert knowledge.
54.5
61.5 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
37.2
—
MathMath problems.
68.3 best
60.4

Benchmarks

30 shared
Only Claude Haiku 5.5: 54Only Inkling 2 Small: 28
Claude Haiku 5.5
Inkling 2 Small
Artificial Analysis Humanity's Last ExamKnowledge
44.4% best28 / 222
44.1%29 / 222
Artificial Analysis Long Context ReasoningReasoning
82.7%23 / 217
83.3% best13 / 217
Artificial Analysis Omniscience AccuracyKnowledge
34.7% best68 / 164
33.8%71 / 164
ARC-AGI-2 (semi-private)Reasoning
61.3%38 / 131
81.7% best20 / 131
ARC-AGI-1 (semi-private)Reasoning
89.0%39 / 126
96.5% best11 / 126
Vibe Code Bench v1.1Coding
90.4% best4 / 123
81.5%28 / 123
ARC-AGI-2 (public eval)Reasoning
64.4%31 / 122
84.2% best20 / 122
ARC-AGI-1 (public eval)Reasoning
90.1%45 / 116
97.4% best20 / 116
Artificial Analysis SciCodeCoding
55.0% best29 / 114
52.3%41 / 114
Legal Research BenchLegal
43.8%25 / 90
46.6% best18 / 90
Tax Agent BenchFinance
31.9%27 / 82
38.9% best11 / 82
ProgramBenchCoding
2.5% best8 / 65
0.5%19 / 65
Terminal-Bench 4.0Agentic
36.4% best21 / 61
12.6%41 / 61
Agent Arena steerabilityAgentic
2.424 / 54
9.3 best13 / 54
Agent Arena task outcomeAgentic
1.628 / 54
3.3 best23 / 54
Humanity's Last Exam without toolsKnowledge
45.9% best14 / 53
44.1%17 / 53
CyberBench v1.1Other
75.5% best6 / 48
72.7%16 / 48
Terminal-Bench-ScienceScience
10.0% best16 / 44
2.9%33 / 44
SRE BenchOther
5.7% best16 / 36
1.9%25 / 36
HLAB-Vals-1.1Legal
4.2%30 / 35
21.7% best6 / 35