vs
5:2
overall capability wins (0 ties across 7 scored). Hy4 Preview leads overall.
- Reasoning1:0
- Coding3:0
- Tool use1:0
- Pricing3:0
- Specs0:2
Verdict
Hy4 Preview leads coding (widest gap: +214.3 on WebDev Elo); Inkling offers larger context (1.0M vs 1M); Hy4 Preview is ~1.4x cheaper on a blended token basis than Inkling.
- Hy4 Preview leads coding (widest gap: +214.3 on WebDev Elo)
- Inkling offers larger context (1.0M vs 1M)
- Hy4 Preview is ~1.4x cheaper on a blended token basis than Inkling
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Hy4 Preview
- $1.46
- Inkling
- $2.01
Hy4 Preview is about 1.4x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
Tencent
Thinking Machines
Family
Other
Other
License
Apache 2.0
Apache 2.0
Open weights
Yes
Yes
Release
Aug 28, 2026
Jul 15, 2026
Knowledge cutoff
-
-
API / provider
Tencent Cloud TokenHub
OpenRouter / Tinker (ref.)
Modalities
text → text
text, image, audio → text
Specs
Context window
1M
1.0M
Max output
-
131K
Parameters
770B (49B act.)
975B (41B act.)
Pricing
Input $/1M
$0.83
$1
Output $/1M
$2.50
$4.05
Blended $/1M (3∶1)
$1.25
$1.76
Speed
tok/s
-
55
TTFT (s)
-
0.55
Reasoning
MMLU-Pro
—
—
GPQA Diamond
92.3
87.2
Humanity's Last Exam
—
31.9
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
41.6
AA-LCR v1.1
—
77.3
CritPt
—
5.4
MMMU-Pro
—
73.5
IFBench
—
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
77.6
SWE-bench Pro
65.7
54.3
SWE-bench Multilingual
82.9
—
LiveCodeBench
—
—
Terminal-Bench 2.1
85.4
55.1
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
—
47.0
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
64.3
—
WebDev Arena
1,624
1,410
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
1,440
| Metric | Hy4 Preview | Inkling | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Tencent | Thinking Machines | — |
| Family | Other | Other | — |
| License | Apache 2.0 | Apache 2.0 | — |
| Open weights | Yes | Yes | — |
| Release | Aug 28, 2026 | Jul 15, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Tencent Cloud TokenHub | OpenRouter / Tinker (ref.) | — |
| Modalities | text → text | text, image, audio → text | — |
| Specs | |||
| Context window | 1M | 1.0M | B +49K |
| Max output | - | 131K | — |
| Parameters | 770B (49B act.) | 975B (41B act.) | B +205 |
| Pricing | |||
| Input $/1M | $0.83 | $1 | A +$0.17 |
| Output $/1M | $2.50 | $4.05 | A +$1.55 |
| Blended $/1M (3∶1) | $1.25 | $1.76 | A +$0.51 |
| Speed | |||
| tok/s | - | 55 | — |
| TTFT (s) | - | 0.55 | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 92.3 | 87.2 | A +5.1 pts |
| Humanity's Last Exam | — | 31.9 | — |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | 41.6 | — |
| AA-LCR v1.1 | — | 77.3 | — |
| CritPt | — | 5.4 | — |
| MMMU-Pro | — | 73.5 | — |
| IFBench | — | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | 77.6 | — |
| SWE-bench Pro | 65.7 | 54.3 | A +11.4 pts |
| SWE-bench Multilingual | 82.9 | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 85.4 | 55.1 | A +30.3 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | — | 47.0 | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | 64.3 | — | — |
| WebDev Arena | 1,624 | 1,410 | A +214 Elo |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | 1,440 | — |
Hy4 Preview
- Context
- 1M
- Parameters
- 770B (49B act.)
- Price
- $0.83 / $2.50
- Speed
- —
- Modalities
- text → text
- License
- Apache 2.0
Tencent Hy Team's Aug 28, 2026 open-weight flagship (770B MoE / 49B active, Apache 2.0) with a 1M context window and default high reasoning effort. Text-only instruct weights on Hugging Face; hosted API at $0.834/$2.501 per 1M tokens via TokenHub and OpenRouter.
Inkling
- Context
- 1.0M / 131K out
- Parameters
- 975B (41B act.)
- Price
- $1 / $4.05
- Speed
- 55 tok/s · 0.55s TTFT
- Modalities
- text, image, audio → text
- License
- Apache 2.0
Thinking Machines' first open-weights MoE (975B / 41B active) — natively multimodal with variable thinking effort and 1M context; Apache 2.0 weights on Hugging Face.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.