Inkling Small
Thinking MachinesReleased Jul 30, 2026Open weights · Apache 2.0inkling-small
- Intelligence (Epoch)
- 150 · 39th of 109
- Reasoning
- yes, with effort levels
- Context
- 1M tokens
- Max output
- 256k tokens
- Input
- text, image, audio
- API features
- tools, seed
Read Nov 3, 11:57 UTC from the hosts' published offers and OpenRouter, spec sheet from Thinking Machines
Free right now
- Free OpenRouter, served by Thinking Machines · until Nov 4 · 20 req/min, 50 req/day, 1,000 req/day with $10 of credits
Checked Nov 3 · every free model on Deals
The cheapest host, DeepInfra, serves fp8 weights (maker: bf16) and serves a 512K context (maker: 1M). Weights and context per host
| Provider | Input | Output | Cache ↑ | tok/s | First token | Uptime 24h | Checked |
|---|---|---|---|---|---|---|---|
| DeepInfrafp8 (bf16) · 512K ctx (1M) · 256K out | $0.45 | $1.20 | $0.10 | 116 | 0.43 s | 100% | Nov 3 |
| BaseTenprecision not stated · 1M ctx | $0.50 | $1.20 | $0.10 | — | — | — | Sep 25 |
2 of 2 providers · prices per million tokens, read Nov 3, 11:57 UTC · speed: OpenRouter, median of the last 30 min, read Nov 3, 11:57 UTC · uptime: OpenRouter, last 24 h · under each host: weights precision, context and max output as served, the maker's value in brackets when the host's differs
| Provider | Weights | Context | Max output | Tools | Structured outputs | Reasoning |
|---|---|---|---|---|---|---|
| DeepInfra | fp8 | 512K | 256K | yes | no | yes |
| BaseTen | not stated by the host | 1M | not stated by the host | — | — | — |
Weights, context, max output: from the host where its terms allow, else as declared to OpenRouter (hover for source and date) · features as declared to OpenRouter, not verified by us
Inkling Small spec sheet, field by field, with sources
Inkling Small spec sheetmaker-declared
What Thinking Machines declares, from its own weights and docs. What each host serves is in the offers: both can be true.
- Context window
- not published by the maker
- Parameters
- 265,956,439,090≈ 266.0B
- Active experts
- 6 of 256 active per token
- Weight precision
- bfloat16
- License
- Apache 2.0 — open weights, permissivecurated · read 2026-09-16Scope: the license of the published weights only — not the maker's acceptable-use policy, nor the terms of the host you call.
- Knowledge cutoff
- to verify
- Modalities
- text · image · audio
- Max output (official)
- not published by the makermaker's docs · read 2026-10-23
- Release date
- to verify≈ July 30, 2026 third-party catalogue — not the maker's figure
- Maker lifecycle status
- to verify
Also declared in the weights
- Architecture class
- InklingForConditionalGeneration
- Layers
- 42
- Vocabulary
- 201,024 tokens
- Weights first published
- July 27, 2026
- Weights last updated
- July 31, 2026
thinkingmachines/Inkling-Small · config.json @ 8cc5877 · read 2026-10-08 · 5 of 10 fields published by the maker
| Provider | Chatbot 1,000 in / 500 out | RAG / long context 8,000 in / 500 out | Batch summarize 4,000 in / 1,000 out |
|---|---|---|---|
| DeepInfra | $1.05 | $4.20 | $3.00 |
| BaseTen | $1.10 | $4.60 | $3.20 |
What each host bills for 1,000 requests of a representative job
Your own workload
No longer offered
These hosts no longer appear where their price is read, so they are left out of every comparison on this page. The last price read is kept for the record.
No longer offered (1)
- Together AI · last price $0.50 in / $1.20 out, read Sep 22 · withdrawn by the host
Intelligence for the price
9 models score higher for the same price or less; it ranks 10 of the 48 models at its price or below.
Epoch Capabilities Index against the lowest input + output price per M, log scale · prices Nov 3. Score: Epoch AI, CC BY 4.0, 150 with an interval of 148 to 152.
Against its rivals
The best of its price range and the top score
- Inkling Smallthis model$1.65150
- GPT-6 Astratop score$60.00169
- GPT-5.6 Lunasame price range$1.40157
- DeepSeek V4 Pro 0813same price range$2.50156
Same level for 6.3× less: GLM-5.3-Flash at Pareto Inference, $0.26 per M input + output · their Epoch intervals overlap (153 against 150).
Bars: Epoch Capabilities Index, drawn from 100 · price: lowest input + output per M.
Run it yourself
open weights, Apache 2.0 · fits 2 of the 14 workstations in our guideSelf-host guide ›- Parameters
- 276B
- 12B active per token (MoE)
- Memory needed
- 166 GB
- in INT4 · 331 GB in INT8
- Smallest cloud setup
- 4× H100
- 320 GB · $10.56/h, about $7,709 a month
- Worth it above
- 5B tokens
- a month, against the cheapest API at $1.65 per M
| Quantization | Memory | Smallest cloud GPU | Quality |
|---|---|---|---|
| INT4 (Q4) | 166 GB | 4× H100 · $10.56/h | smallest footprint, minor quality loss |
| INT8 (Q8) | 331 GB | 8× H100 · $21.12/h | near-lossless |
| FP16 (full) | 662 GB | multi-node, over 640 GB | reference quality |
At home: in INT4 it fits the Mac Studio M5 Ultra (256 GB unified) — about $12,299 once, or $0.468/h over 3 years, electricity aside, at 1200 GB/s of memory bandwidth — that, not FLOPS, sets tokens per second; it pays for itself after about 7B tokens against the cheapest API.
Memory ≈ parameters × bytes × 1.2 · cloud: on-demand, 24/7 · builds: GGUF, AWQ, MLX, GPTQ, Ollama · break-even calculator
Other Thinking Machines models · all its models, hosts and history ›
- Inkling3 providers · from $5.00