Atria Dawn Preview vs Claude Haiku 5.5
Shanghai AI Laboratory · China | Anthropic · US · Updated June 2026
Quick verdict
Pick Atria Dawn Preview for best-in-class among tested models on several of shanghai ai lab's own benchmarks, including browsecomp (92.5, ahead of gpt-5.6 sol's 92.2) and deepsearchqa (96.0) or 744b-parameter moe built on a glm-5.2 base, released under a fully open mit license. Pick Claude Haiku 5.5 for 5x the context window of haiku 4.5 (1m vs 200k tokens) and double the max output (128k vs 64k) or first haiku-class model with an adjustable 'effort' setting to trade off cost, latency, and intelligence. Choose Atria Dawn Preview if you need self-hosting or data privacy; Claude Haiku 5.5 if you want a managed API.
Atria Dawn Preview (Shanghai AI Laboratory, China) and Claude Haiku 5.5 (Anthropic, US) line up two different AI ecosystems against each other — a comparison that is as much about cost philosophy and openness as raw capability. Atria Dawn Preview is a free, MIT-licensed 744B open-weight model from Shanghai AI Lab that beat GPT-5.6 Sol on the BrowseComp benchmark — released quietly, weights before paper. Claude Haiku 5.5 is anthropic's October 7, 2026 small-model refresh — a 1M-token context window (up from 200K) at roughly 75% lower average cost than Haiku 4.5, priced at $0.10/$0.50 per million tokens. They diverge most on price, context window and open vs. closed weights — each quantified below from the models' real specs.
Key differences at a glance
- ▸Cost model: Atria Dawn Preview ships open weights you can self-host (hardware cost only, no per-token fee), while Claude Haiku 5.5 is API-metered at $0.1/$0.5 per 1M tokens. Your choice depends on whether you want zero marginal cost at the price of running infrastructure.
- ▸Context window: Claude Haiku 5.5 holds 3.8× more — 1M tokens (~1,500 pages) vs 256K tokens (~393 pages). But effective recall usually fades long before the advertised ceiling, so the bigger number only helps if the model reasons over it.
- ▸Recency: Claude Haiku 5.5 is the newer model by about 26 days (released October 7, 2026), usually meaning fresher training data and capabilities.
- ▸Ecosystem: this is a China-vs-US matchup — they differ in pricing philosophy, data-residency options, and tooling ecosystems, not only benchmarks.
Side-by-side specs
| Spec | Atria Dawn Preview | Claude Haiku 5.5 |
|---|---|---|
| Provider | Shanghai AI Laboratory (China) | Anthropic (US) |
| Released | September 11, 2026 | October 7, 2026 |
| Context window | 256K tokens (~393 pages) | 1M tokens (~1,500 pages) |
| Price (in/out) | Open weight (self-host / free) | $0.1/$0.5 per 1M tokens |
| Open weight? | Yes — self-hostable | No — API only |
| Modalities | text | text, image, code |
| SWE-Bench Verified | Not published | Not published |
| MRCR v2 @ 1M | Not published | Not published |
Who wins what
Best-in-class among tested models on several of Shanghai AI Lab's own benchmarks, including BrowseComp (92.5, ahead of GPT-5.6 Sol's 92.2) and DeepSearchQA (96.0)
Atria Dawn Preview
A free, MIT-licensed 744B open-weight model from Shanghai AI Lab that beat GPT-5.6 Sol on the BrowseComp benchmark — released quietly, weights before paper — and its weights are open while Claude Haiku 5.5 is API-only.
744B-parameter MoE built on a GLM-5.2 base, released under a fully open MIT license
Atria Dawn Preview
Open weights make this possible at all — Claude Haiku 5.5 is API-only, so it cannot leave the vendor's servers.
Free to self-host — no API pricing, run entirely on your own hardware
Atria Dawn Preview
Claude Haiku 5.5 is comparatively weak here — pricing jumps roughly 5x beyond 100K-token prompts ($0.50/$2.50 per million vs the headline $0.10/$0.50)
5x the context window of Haiku 4.5 (1M vs 200K tokens) and double the max output (128K vs 64K)
Claude Haiku 5.5
Its 1M tokens window holds about 3.8× more than Atria Dawn Preview's 256K tokens in a single prompt.
First Haiku-class model with an adjustable 'effort' setting to trade off cost, latency, and intelligence
Claude Haiku 5.5
Anthropic's October 7, 2026 small-model refresh — a 1M-token context window (up from 200K) at roughly 75% lower average cost than Haiku 4.5, priced at $0.10/$0.50 per million tokens — and it carries the larger 1M tokens context.
Anthropic's cheapest small model yet — runs about 75% cheaper on average than Haiku 4.5
Claude Haiku 5.5
Anthropic's October 7, 2026 small-model refresh — a 1M-token context window (up from 200K) at roughly 75% lower average cost than Haiku 4.5, priced at $0.10/$0.50 per million tokens — and it is the newer of the two.
Lowest cost at scale
Atria Dawn Preview
Its weights are open, so at volume you pay for your own hardware instead of Claude Haiku 5.5's $0.1/$0.5 per 1M tokens.
Largest single-prompt input
Claude Haiku 5.5
Its 1M tokens window is about 3.8× larger than Atria Dawn Preview's 256K tokens, fitting roughly 1,500 pages in one prompt.
Which should you pick?
A cost-sensitive startup shipping high volume
→ Atria Dawn Preview
At Open weight (self-host / free) it undercuts Claude Haiku 5.5, and on millions of tokens that margin decides the monthly bill.
Someone analysing very long documents or codebases
→ Claude Haiku 5.5
Larger 1M tokens window fits more in one prompt.
A team with data-privacy or self-hosting needs
→ Atria Dawn Preview
Open weights let you run it on your own hardware; Claude Haiku 5.5 is API-only.
Anyone whose priority is best-in-class among tested models on several of shanghai ai lab's own benchmarks, including browsecomp (92.5, ahead of gpt-5.6 sol's 92.2) and deepsearchqa (96.0)
→ Atria Dawn Preview
It is specifically built for that.
Anyone whose priority is 5x the context window of haiku 4.5 (1m vs 200k tokens) and double the max output (128k vs 64k)
→ Claude Haiku 5.5
That is its strongest area.
An enterprise with regional data-residency rules
→ Claude Haiku 5.5 or Atria Dawn Preview
Origin (China vs US) affects where data is processed and which compliance regime applies — check the provider's terms for your region.
Atria Dawn Preview: where it fits
A free, MIT-licensed 744B open-weight model from Shanghai AI Lab that beat GPT-5.6 Sol on the BrowseComp benchmark — released quietly, weights before paper. Released September 11, 2026 by Shanghai AI Laboratory, it is built for best-in-class among tested models on several of Shanghai AI Lab's own benchmarks, including BrowseComp (92.5, ahead of GPT-5.6 Sol's 92.2) and DeepSearchQA (96.0), 744B-parameter MoE built on a GLM-5.2 base, released under a fully open MIT license, and free to self-host — no API pricing, run entirely on your own hardware.
Its trade-offs are real: context window is reported inconsistently across trackers — most list 256K tokens, at least one lists 1M; unconfirmed which is accurate, released quietly, with weights and an API posted to GitHub before Shanghai AI Lab's own technical report went up on arXiv days later, and a lab research preview rather than a commercial product — support and update cadence are unclear. As an open-weight model, its running cost is your own hardware rather than a per-token fee.
Claude Haiku 5.5: where it fits
Anthropic's October 7, 2026 small-model refresh — a 1M-token context window (up from 200K) at roughly 75% lower average cost than Haiku 4.5, priced at $0.10/$0.50 per million tokens. Released October 7, 2026 by Anthropic, it is built for 5x the context window of Haiku 4.5 (1M vs 200K tokens) and double the max output (128K vs 64K), first Haiku-class model with an adjustable 'effort' setting to trade off cost, latency, and intelligence, anthropic's cheapest small model yet — runs about 75% cheaper on average than Haiku 4.5, and over-refuses benign requests only 0.17% of the time on the API, down from Haiku 4.5's 0.44%.
Its trade-offs: pricing jumps roughly 5x beyond 100K-token prompts ($0.50/$2.50 per million vs the headline $0.10/$0.50), extended thinking can no longer be fully disabled — only dialed between low, medium, and high effort, and as the smallest tier, trails both Sonnet 5.5 and Opus 5.5 on hard reasoning and open-ended work. At $0.1 in / $0.5 out per million tokens, it sits in the budget price band.
The bottom line for this matchup
The defining split here is open vs. closed. Atria Dawn Preview gives you weights you control — self-host it, fine-tune it, keep data in-house, pay only for hardware. Claude Haiku 5.5 gives you a managed, always-updated API with no infrastructure to run. Teams with GPUs, privacy requirements, or huge volume often favour the open model; teams that want zero ops and the latest capabilities favour the closed one. Capability is close enough that this operational question, not the benchmark, usually decides it.
Want both Atria Dawn Preview and Claude Haiku 5.5 without two subscriptions? LumiChats gives you these plus 40+ models under one ₹69/day pass (about $1/day) — draft with one, cross-check with the other.
See pricingFrequently asked questions
Is Atria Dawn Preview or Claude Haiku 5.5 better for coding?
Public SWE-Bench figures are not available for either model, so the honest test is your own repository — run an identical real bug through both. By design, Atria Dawn Preview leans toward best-in-class among tested models on several of shanghai ai lab's own benchmarks, including browsecomp (92.5, ahead of gpt-5.6 sol's 92.2) and deepsearchqa (96.0) while Claude Haiku 5.5 leans toward 5x the context window of haiku 4.5 (1m vs 200k tokens) and double the max output (128k vs 64k), and that positioning usually predicts which feels better on your codebase.
Which is cheaper, Atria Dawn Preview or Claude Haiku 5.5?
Atria Dawn Preview is open-weight, so self-hosting means no per-token fee (you pay for hardware instead), while Claude Haiku 5.5 is API-metered at $0.1/$0.5 per 1M tokens. For most teams without GPUs, the API model is cheaper to start; at very high volume, self-hosting can win.
Which has the bigger context window?
Claude Haiku 5.5 — 1M tokens vs 256K tokens, about 3.8× larger. Useful only if the model actually reasons over the full window, which not all do.
Can I use both Atria Dawn Preview and Claude Haiku 5.5 together?
Yes — a multi-model platform like LumiChats gives you Atria Dawn Preview, Claude Haiku 5.5 and 40+ others under one ₹69/day pass (about $1/day), so you can draft with one and cross-check with the other instead of buying two subscriptions.
Which is newer, Atria Dawn Preview or Claude Haiku 5.5?
Claude Haiku 5.5 — released October 7, 2026, about 26 days after Atria Dawn Preview.
Related comparisons
Specifications and benchmarks reflect publicly reported figures as of June 2026 and may change as providers release updates. Always verify on your own workload.